Machine Learning Model for Predicting Potential Donors Using Logistic Regression

Natural calamities like hurricanes, tsunami and pandemic are tend to happen so often in today’s world. Under this scenario, predictive modelling is helpful in terms of resources allocation to achieve the objective effectively. This study intends to construct a prediction model based on logistic regression to predict the possible donors who can help in such tragic situations. Sample dataset is taken from internet source. Initial data exploration being performed to better understand the variables in dataset. To improve the quality of dataset, missing value treatment and feature engineering are performed before the construction of prediction model. During the missing value treatment, various methods being applied with mean imputation has the better performance in terms of variable significance and standard error. Feature engineering including one-hot encoding, categorical grouping, multicollinearity treatment and log transformation being performed. During the modelling phase, normal logistic regression and stepwise logistic regression being performed. The performance of the models was measured by Accuracy, Sensitivity and Specificity of the training and testing dataset. The Stepwise Logistic Regression outperformed the normal Logistic Regression with model accuracy at 58.5% along with sensitivity rate of 54.3% and specificity rate of 62.6%Keywords—Machine Learning, Recommender System & Feature Extraction.