Promedio Movil Simple

How to Calculate Forecasts Using Simple Average Techniques

Initial Observations on Forecasting Methods

• The method discussed does not consider all historical data; it focuses on the most recent data points for forecasting.

• For example, to forecast January sales, only December and November data are used, rather than all historical data.

Moving Averages Explained

• A moving average can be calculated over 'n' periods, where 'n' represents the number of months or years considered in the average.

• The moving average is defined as the arithmetic mean of the most recent 'n' periods. It helps predict only for the next period based on available real information.

Limitations of Moving Averages

• The technique cannot forecast beyond one period ahead; if data is available only until October, predictions can only be made for November.

Practical Example: Sales Forecasting for 2018

• An example involves a company wanting to forecast sales for 2018 using historical sales data from previous years (2006 to 2017).

• The initial focus is on calculating forecasts using three different values of 'n': n3, n4, and n5.

Step-by-Step Calculation Process

• To forecast for 2009, three prior years (2008, 2007, and 2006) are needed. If not enough historical data exists, cells remain empty.

• For each subsequent year (e.g., 2012), averages are calculated by summing previous years’ sales figures and dividing by 'n'.

Results from Calculations

• Following this process yields a forecast of approximately 24 units for 2012 when using an n value of 3.

• Continuing with calculations leads to a final forecast of about 41.33 units for the year 2018.

Exploring Different Values of 'n'

• By changing 'n' to equal four instead of three and repeating calculations with available historical data results in different forecasts.

• This highlights that varying 'n' affects outcomes significantly; thus careful consideration is necessary when selecting it.

Conclusion on Forecasting Techniques

• There are distinct differences in forecasts depending on whether an n value of three or four is used; understanding these variations is crucial in forecasting accuracy.

Forecasting Analysis with Historical Data

Understanding the Forecasting Process

• The speaker discusses setting n equal to 5, indicating that five previous data points will be used for forecasting. This involves removing the oldest data point to maintain a consistent dataset.

• The calculation of forecasts is explained by summing data from 2009 to 2013, emphasizing the importance of having sufficient historical data for accurate predictions.

• A forecast for the year 2018 is presented, showing different results based on varying values of n. For instance, with n set to 5, the forecast is noted as 37.

Comparing Different Forecasts

• The speaker notes that using n equal to 3 yields a forecast of approximately 41. In contrast, using n equal to 4 gives a lower forecast of about 39.

• The discussion raises an important question: which forecast should be considered the best? It suggests evaluating forecasts based on their accuracy and proximity to historical data.

Evaluating Forecast Accuracy

• To assess accuracy, the concept of error is introduced. Error measures how far off a forecast is from actual historical data.

• A visual representation (graphical analysis) will be created using forecasts from different values of n, allowing for easier comparison against historical demand levels.

Graphical Representation and Insights

• When analyzing graphical trends, it becomes evident that different values of n yield varying distances from actual historical data.

• The conclusion drawn from graphical analysis indicates that an n value of 3 provides the closest approximation to historical demand levels, suggesting it may be the most reliable forecast method.

Mathematical Calculation of Errors

• The mathematical approach involves calculating errors by subtracting predicted values from actual demand figures.

• An example illustrates this process: if demand equals 30 and prediction equals 24, then the error can be calculated directly.

• Squaring these differences helps eliminate negative values in error calculations since both overestimations and underestimations are treated equally in terms of distance from actual figures.

Summarizing Total Error

• After squaring individual errors, a total sum is computed to determine overall forecasting accuracy across all periods analyzed.

Understanding Error Summation in Forecasting

Historical Context of Data Summation

• The process involves taking the summation of errors historically, which is crucial for forecasting accuracy.

• The average of squared errors is calculated to determine the best forecast; lower averages indicate better predictions.

Calculation Methodology

• To find the average error, sum the squared differences between actual demand and forecasts, then divide by the number of data points.

• Errors are computed as the difference between demand and forecast values, followed by squaring these differences for analysis.

Data Analysis Steps

• After calculating individual errors, their sum is divided by the total number of data points to find an average error.

• The analysis excludes certain data (like risk factors), focusing solely on available historical forecasts for accurate comparisons.

Comparative Results

• For n = 4, an error value of 51 was found; for n = 5, it was 75.37. This comparison helps identify which model performs better.

• Graphical representation confirms that n = 3 yields the lowest error, making it the most reliable forecast compared to others.

Conclusion on Best Forecasting Technique