Time series forecasting using Regression Analysis in R

Question

As part of my work, I need to evaluate different forecasting models on the time series data using R and pick the one with lowest error. For this, I want to know how to use the Linear Regression(LR) method to forecast on the time series. In the time series, we normally have only 1 column with continuous data but to use LR, we need at least 2 variables, like y=Beta0+Beta1*x. I have the sales figure monthly(x) but how to get the y variable to use LR.

Answer 1

Praveen, I don't think we are focusing the problem right here.I understand you are trying to forecast sales in the future based on past data.

Let's suppose you have defined a data.frame sales , with revenue and time_period data.

At this moment, I would personally do some data exploration to check there might be some kind of correlation between the variables.

xyplot(revenue ~ time_period, data = sales)

Then, fit the model

sales_model = lm(revenue ~ time_period, data = sales)

Now the model is built and the coefficients you are looking for should be calculated. To get them, just run summary(sales_model)

The first column of the coefficients matrix should give you both the intercept and the slope of your model, which are the betas you were looking for.