Skip to content

Classification

Introduction

TIM performs binary classification if you provide your target variable as a Boolean with zeroes and ones. Classification of more than 2 classes is currently not available. TIM behaves the same way as for any other dataset including feature engineering in the expansion phase, reduction and multiple model creation.

Dataset format

Your classification dataset has therefore the same format even if you do not have time dependencies in your data. In this case you can simply use indices from 1 to n instead of timestamps.

image.png

Classification and Forecasting Likelihood of Being Inside a Corridor

In cases where your target variable is only a reflection of another continuous variable - e.g. you would like to forecast if this variable stays inside of a given corridor or not - you could still provide this variable among predictors, but make sure its availability is set properly so the model does not learn this transformation (e.g. "corridor" transformation) instead.

image.png

Every forecast can be interpreted as a likelihood of the actual value being 1 - e.g. forecast 0.6 could be interpreted as a 60 percent chance for the record belonging to the class 1 and 40 percent for it belonging to the class 0.