3. Time Series Analysis with Machine Learning

Accessing and preprocessing data

Read website.txt file to get daily numbers of website visitors. Note: "Support short data rows" must be enabled.

Detect & deal with seasonality

Rebuilding the signal

Workflow: Time Series Analysis with Machine Learning

This workflow shows an example of a time series analysis using the verified components Inspect Seasonality and Remove Seasonality.

The data for this example is website traffic data, reflecting the daily number of website visitors, and we want to predict the next day's number of visitors given the number of visitors in the past n days. However, to detect and deal with the seasonality in the data we first use the Inspect Seasonality component to find a sign of dominant seasonality. Then, to remove seasonality trends in the data, we use the Remove Seasonality components. In this case we define n (the lag value) as 7. Lastly, we train and apply a Linear Regression, however, any other numerical prediction algorithm would have worked as well.

Create lag intervals

Copy and shift "no visitors" column n times: from x(t) to x(t), x(t-1), x(t-2), ..., x(t-n).

n = 7 (lag value)

Training Linear Regression & applying trained model on test set

Evaluate prediction