You are here:
R2 Quality Alert
R2 (R-squared), or the coefficient of determination, is a performance metric that evaluates how well a regression model fits the data.
Actions to Consider
- A low R2 indicates that the model doesn’t perform well. Improve the model before you activate it.
- A high R2 indicates that the model performs so well that data leakage can be a factor. Investigate and remove any variables that cause data leakage.
Detection Methodology
Model Builder displays an alert when it detects a high or low R2 score based on these values.
- Below 0.3 (low)
- Above 0.95 (high)
Example
A retail brand wants to predict monthly sales for each store. To achieve this, the retailer builds a regression model with these input variables.
- store size (sq. ft)
- number of employees
- average customer footfall per day
- advertising spend per month
- local population density
After model training, an alert displays because the R-squared score is 0.45, which is considered low. The score indicates that the model explains only 45% of variation in the "monthly sales" target variable and that it isn’t capturing other factors that are influencing sales. Thus, the predictions can be unreliable. To resolve the issue, here are some actions to consider.
- Improve model performance with additional variables such as “store promotion” or “competitor proximity".
- Check the data quality to identify missing values, outliers, and inconsistencies.
- Use more historical sales data to improve the quality of data.
- Use more complex modeling algorithms, such as random forest or gradient boosting, if linear models underperform.
- Compare the difference between actual and predicted sales to uncover patterns or inconsistencies.
- Retrain the model with an updated dataset.
