The Impact of Feature Selection on the Regression Task for Life Expectancy Prediction

Life expectancy prediction is important for decision-making to facilitate advanced care planning processes. This is possible by applying the needed treatment to foresee the procurement of health care services. Machine learning techniques are used to predict the life expectancy values for different people using regression algorithms. Feature selection methods are known for their power to enhance the prediction quality results by selecting the best features that affect the prediction of the results. In this paper, different regression techniques are applied to a life expectancy dataset provided by the World Health Organization. These algorithms include Linear Regression, K-Nearest Neighbor Regression, and Decision Tree Regression which are combined with two different feature selection methods which are the Correlation Feature Selection and The Mutual Information Feature Selection methods. Moreover, different k values are experimented with for the feature selection methods to generate the best results in terms of R
²
and RMSE evaluation measures. The results show that the Decision Tree Regression method produced the best results for the two evaluation measures with a combination of the Mutual Information Feature Selection method with a k value of 15 and the Correlation Feature Selection method with a k value of 10 for the R
²
and RMSE, respectively.