An Effective Stroke Prediction Based on Cat Boost with Few Short Compound Activity

In recent years, strokes are becoming a maj or global health concern that affecting the welfare of the population. Many researches have focused on developing predictive models using Machine Learning (ML) and Deep Learning (DL) for strokes. However, the existing method Support Vector Machine (SVM) used to predict the strokes but struggled with sensitivity issues. To overcome this issue, Cat Boost with Few-Shot Compound Activity Prediction (CB-FSCAP) is proposed for stroke prediction. The Few-Shot Compound Activity Prediction (FSCAP) effectively learns the representation from labeled data to reduce the sensitivity issues. Primarily, the input data is taken from healthcare-dataset-stroke-data dataset and the further processed into preprocessing. The stop words are removed and imputation, lemmatization are utilized to reduce dimensionality in preprocessing. Then, the preprocessed data is fed into Cat Boost (CB) for selecting the optimal features. Finally, prediction is done with the help of FSCAP to predict the strokes in various emergency conditions. From the results, the Proposed CB-FSCAP established better results when compared to existing Random Forest (RF) by providing results in terms of accuracy (98.4%), precision (95.1 %), recall (96.50/0) and F1-score (95.7%) respectively.