Classification of Student Mental Health Prediction Using LSTM

A novel approach for identifying the metal health of the students that are indicative of depression due to Covid-19 Pandemic, In this study Natural Language Processing techniques such as embedding of words, tokenizing, and vectorization to are used extract sentiment from the data that emotions of sadness, depression, and similar emotions. The prediction of sentiment in the text is made accurately using Deep Learning, that eliminates false positives using the immediate context of words. Dataset for the study has been collected for over more than 2 lakhs tweets and posts from public forums and social media website for training the model developed. Labelling is done before data analyzing, which permits posts that contain common topics are grouped together. data from many groups expressing about depression and sadness are considered as the positive class, where as the posts from any random sites, miscellaneous are considered as the negative class. The dataset may be said to be indicative of a real-world environment due to the diversity of the negative class. The model developed in this study has applications across a broad spectrum of domains, such as deployment on online platforms and communication forums used by students to detect potentially harmful tendencies. This study helps various educational institutes, private and governments organisations in analysing student and public sentiments, as well as their attitude in response to the pandemic, and then take precautions and actions accordingly, the algorithm that used here is LSTM, In this study it is observed that LSTM provides accuracy of 88%.