Questions to Answer - Linear Regression: 1) What is the ideal window for the linear regression technique (from the below

[answerhappygod]

Questions To Answer Linear Regression 1 What Is The Ideal Window For The Linear Regression Technique From The Below 1 (246.19 KiB) Viewed 46 times

Questions to Answer - Linear Regression: 1) What is the ideal window for the linear regression technique (from the below list)? Please create a graph that compares the average forecasting performance for all ten installations against the window size. Choose from among window sizes in minutes) of [2,5,10,15,30,60,120). Questions to Answer – ARIMA: 2) To select parameters for the p and q parameters of the ARIMA model, you should produce a graph of the autocorrelation function (ACF) and partial autocorrelation function (PACF) for each of the ten solar installations. From these plots, determine the values of p and q by seeing which values lie outside the significance region (see [2] above for a demonstration). To answer this question, explain how you chose the ideal values of p and q, and include at least one plot of the ACF and PACF in your write-up (you do not need to include all of the plots you generate). Questions to Answer - Performance Comparison: 3) In class, we discussed three different metrics for assessing errors in a forecasting problem: Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE). Please create three graphs (one for each metric) that compare the performance of the three forecasting models (persistence, linear regression, and ARIMA) averaged over the 10 installations. You should use the ideal parameters for the models that you found in the earlier problems. Considering the three metrics (RMSE, MAE, and MAPE) - which is/are best for comparing the performance of these techniques? Please provide reasons for your selection. 4) What differences did you find among the performance of the algorithms? What do you think causes those differences? 5) How does each of these three techniques perform as the resolution of the input data changes? Please create a graph that shows the performance of each technique averaged over the ten installations with data resolutions in minutes) of (1,5,15,30,60). Note that in order to create a lower resolution dataset, you should sample from the dataset, rather than averaging together values. For linear regression, choose a window size that matches the duration of history you found in question 1 (note that this will change the number of samples your window will include as the input data resolution changes).