Enhancing Predictive Model Performance through Comprehensive Pre-processing and Hybrid Feature Selection: A Study using SVM

This study aims to achieve the improved performance of heart disease prediction based on the increased level of preprocessing and rather effective feature selection approaches. Starting with data preparation in which the data is first cleaned up the work develops a class balance solution using Synthetic Minority Over-sampling Technique (SMOTE) while the Z-score function identifies outliers. As it will be described in CHAN section that applies Chi-square and ANOVA under a hybrid model, the useful characteristics for model training are identified and prioritized. Using a method derived from the medical records database, the study focuses on the heart health indices using Support Vector Machines (SVM) which is a modelling technique widely known for its effectiveness in handling increased numbers of variables and the relationship between the measurements. Measures such as accuracy, precision, recall and F1-score among others are used in the evaluation of the success of the approach. In particular, the results shed the light to Health Care applications where risk assessment and correct diagnosis is extremely critical; the findings also assert the important roles of strong preprocessing and selective feature selection strategies in enhancing the predictive model’s performance. This work contributes to the development of approaches intended for stable and efficient modelling in medical decision support system.