Data Description:

This data set has 398 observations and contains the variables Date and IPG2211A2N. The data set outlines monthly Electric Production in the united states. The records were taken from 1985 to 2018.

Training and testing data:

The code defines the training and testing datasets for time series forecasting by holding up the last 6 months of the data for calculating forecasting errors. This sets up the data for time series forecasting, allowing different forecasting models to be fitted to the training dataset and evaluated on the testing dataset to measure their accuracy.

Building 4 Forecasting Models:

This uses four baseline forecasting methods and the first 30 data values to forecast the next 6 month’s data values.

The output shows the predicted values for each model for the 15 periods ahead of the time series. The first column represents the predicted values using the mean model, the second column shows the predicted values using the naive model, the third column represents the predicted values using the seasonal naive model, and the fourth column shows the predicted values using the random walk with drift model.

Forecasting Table
pred.mv pred.naive pred.snaive pred.rwf
88.2564 91.4867 93.0628 91.53652
88.2564 91.4867 103.2203 91.58634
88.2564 91.4867 117.0837 91.63616
88.2564 91.4867 106.6688 91.68598
88.2564 91.4867 95.3548 91.73580
88.2564 91.4867 89.3254 91.78562
88.2564 91.4867 90.7369 91.83544
88.2564 91.4867 104.0375 91.88526
88.2564 91.4867 114.5397 91.93508
88.2564 91.4867 115.5159 91.98490
88.2564 91.4867 102.7637 92.03472
88.2564 91.4867 91.4867 92.08454
88.2564 91.4867 93.0628 92.13436
88.2564 91.4867 103.2203 92.18418
88.2564 91.4867 117.0837 92.23400

Visulaization:

A time series plot is now created with the predicted values. Looking at the graph the seasonal naive method performed the best, with moving average, naive, and drift performing poorly comparitavely.

Accuracy Metrics:

This table shows the overall performance of four forecasting methods based on three error metrics: Mean Absolute Percentage Error (MAPE), Mean Absolute Deviation (MAD), and Mean Squared Error (MSE). With the four forecasting methods being Moving Average, Naive, Seasonal Naive, and Drift.

Seasonal Naive has the lowest MAPE, MAD, and MSE values, indicating that it is the most accurate method among the four. Moving Average has the highest values for all three error metrics, indicating that it is the least accurate method among the four.

Naive and Drift have intermediate performance, with Naive having the lowest MSE value and Drift having the lowest MAD value. However, when considering all three metrics together, Seasonal Naive appears to be the best forecasting method.

Overall performance of the four forecasting methods
MAPE MAD MSE
Moving Average 14.159986 234.5564 362.50149
Naive 11.491031 192.3692 271.91123
Seasonal Naive 4.428782 72.3599 37.29849
Drift 11.154238 186.9886 260.83043