Comparative Analysis of Classifcation Algorithm for Hair Fall in Survey Based Dataset

Hair is an important part of an individual's personality and it has a key role in building and boosting someone's confidence. Hair fall has become a serious problem these days. Hair loss is very hard to recover. If hair fall is predicted early then its preventive measures can be applied to reduce its effect. Classification Algorithm can be used for early prediction of hair fall. The data used in this paper is of Primary Source. Open-ended questions and Closed-ended questions were used in questionnaire. The values on the various 22 attributes and 1 class Label were collected using a google form. Pre-processing techniques such as label encoding, one hot encoding, Min-Max scaling and correlation plot was used before training. Grid Search cross validation was used on classification algorithms such as SVM, MLP, XGBoost, Random Forest, Logistic Regression and Decision Trees. All six model were compared using various metrics such as Accuracy, Precision, Recall, F1-Score and MCC among which XGBoost achieved highest accuracy of 91.89%. Furthermore, feature importance graph of XGBoost revealed Hair fall rate, Pain/itch, Scalp Infection, Genetic Factors(Father's Hair fall), Hair line patterns, Medication (Diabetes, Depression), Density, sleep patterns are the most significant features for the prediction of hair fall. The model was finally used to predict on unseen data.