Forecasting (10): Moving average forecast
Summary
TLDRThis video explains the concept of moving average models in forecasting. It distinguishes between simple averages and moving averages, where the latter involves calculating averages over a specific number of previous data points (e.g., 2, 3, 10 periods). The presenter demonstrates how to apply this in Excel, comparing different moving average periods and their forecast accuracy. It also covers calculating forecast errors and evaluating the effectiveness of the model using Mean Absolute Percentage Error (MAPE). The video concludes with an introduction to the weighted moving average model.
Takeaways
- ๐ Moving average forecasts use a fixed period (e.g., 2, 3, 10 periods) to calculate the average of the previous data points for each forecast.
- ๐ The moving average formula averages a set number of previous data points (e.g., 3-period, 5-period) to generate forecasts for future periods.
- ๐ In Excel, you can apply the moving average by using the AVERAGE function to calculate the forecast based on the selected period.
- ๐ To calculate the forecast error, the absolute difference between actual values and forecasted values is divided by the actual values, providing a percentage error.
- ๐ When no real data is available for the forecast point, the previous forecast value can be used as a substitute in the moving average calculation.
- ๐ The moving average model can adapt to various time series by experimenting with different periods (e.g., 2-period, 3-period, 10-period) to see which yields the best results.
- ๐ Moving averages differ from static forecast models, as they adjust with each new data point and vary for each period.
- ๐ Comparison of models (e.g., moving average, dynamic forecast) can be done using MAPE (Mean Absolute Percentage Error) to determine which model performs best for a given dataset.
- ๐ A moving average model, when applied with a rolling window, continuously updates the forecast using only the most recent data points.
- ๐ A rolling forecast uses a set of recent data points for predictions, while an expanding forecast gradually increases the data set over time to improve accuracy.
Q & A
What is the main difference between the previous model and the moving average model?
-In the previous model, the total leverage was calculated by averaging all available data points. In the moving average model, a specific period is selected (e.g., 2, 3, 5, 10 periods), and for every forecast point, only the average of the last specified periods is taken into account.
How does the moving average model work in terms of data points?
-In the moving average model, for every forecast point, the average of a set number of previous periods is calculated. For example, with a 3-period moving average, only the last 3 data points are considered for forecasting.
How is the moving average forecast formula represented mathematically?
-The moving average forecast is represented as Yt = (Sum of previous periods) / Number of periods. For example, if using 3 periods, Yt = (Yt-1 + Yt-2 + Yt-3) / 3.
What is the importance of adjusting the moving average period?
-The moving average period can be adjusted (e.g., 2, 3, 5, or 10 periods) to find the most accurate forecast for a given time series. Experimenting with different periods helps determine which one provides the best prediction accuracy.
How is the moving average forecast calculated in Excel?
-In Excel, you can calculate a moving average by taking the average of a specific number of periods, then dragging the formula down the column to apply it to all subsequent forecast points. The forecast adjusts as new data points are added.
What is the error calculation for the moving average forecast in the transcript?
-The error is calculated by finding the absolute difference between the real value and the forecasted value, then dividing by the real value. This provides a percentage error for each period.
What is the mistake noted in the transcript when calculating the moving average forecast?
-The mistake was initially using the average of only available data points, without considering that some forecast periods might have fewer actual data points available. The corrected approach involves using the available data and previous forecast values to calculate the moving average.
How does the moving average forecast differ from the dynamic forecast?
-The moving average forecast results in different values for each forecast period, unlike the dynamic forecast, which produces the same value for all periods. This difference is due to the moving average incorporating the most recent data points in each forecast.
What are the different approaches mentioned for calculating forecast errors?
-The transcript mentions three different forecast approaches: dynamic forecasting, averaging, and moving average, with error calculations performed for each. The moving average method produced varying forecast values, which affected the error outcomes.
What should be done if a longer moving average period is chosen, like 10 periods?
-If a longer moving average period, such as 10 periods, is chosen, the model will always take the average of the last 10 data points for forecasting. This approach works well with large data samples and helps smooth out fluctuations.
Outlines

This section is available to paid users only. Please upgrade to access this part.
Upgrade NowMindmap

This section is available to paid users only. Please upgrade to access this part.
Upgrade NowKeywords

This section is available to paid users only. Please upgrade to access this part.
Upgrade NowHighlights

This section is available to paid users only. Please upgrade to access this part.
Upgrade NowTranscripts

This section is available to paid users only. Please upgrade to access this part.
Upgrade NowBrowse More Related Video

Peramalan Permintaan Part 1 (Moving Average and Weighted Moving Average)

Forecasting (1): What is forecasting?

Introduction to Time Series Forecasting | SCMT 3623

Identification of a Time Series using the ACF and PACF

Exponentially Weighted Moving Average or Exponential Weighted Average | Deep Learning

MANAJEMEN OPERASI - Memahami Peramalan (Forecasting)
5.0 / 5 (0 votes)