Levy Flight Based Gannet Optimization Algorithm with Euclidean Distance-Support Vector Machine for Predictive Analysis of Mental Health Crisis Using Social Media Data

The analysis of mental health crisis helps to know the mindset of the social media users so that the applications are modified based on the user requirements. The existing algorithms failed perform the predictive analysis of mental health crisis using social media data due to lack of dimensionality handling ability, that reduced the performance the model. The LF-GOA with ED-SVM is proposed to overcome the problem of existing algorithms for the mental health crisis predictive analysis. The twitter data is used as input for the predictive analysis model. The data was pre-processed using data cleaning, stemming, tokenization and stop word removal techniques. The n-gram and Term Frequency-Inverse Document Frequency (TF-IDF) were used to get the feature vectors and the Levy Flight based Gannet Optimization Algorithm (LF-GOA) was used for the selection of significant features from the data. At last, the Euclidean Distance -Support Vector Machine (ED-SVM) classifier was used for the prediction of mental health crisis. The performance of the proposed LF-GOA with ED-SVM model is evaluated and obtained better results of 98.79% accuracy, 97.81% precision, 97.45% recall and 97.62% f1-score for sentiment 140 dataset compared to the existing Gated Attention Recurrent Network (GARN) method.