Mental Health has become a very serious issue in today’s life. Excessive of mental pressure leads to suicide whose numbers are drastically increasing. According to report of WHO, India had the highest suicide rate in the South-East Asia. Over 134,000 deaths due to suicides were recorded in India during 2018. All over world on an average 800,000 deaths occur due to suicides. Mental pressure needs to be decreased which in turn reduces suicide. But as per the reports, most of the victims prefer not to share their mental health issue to others. We approach to solve this by: Model will be trained for sentiment analysis. MRI images will be used to study the structure of brain and affects on it so that we predict brain pressure and other mental health diseases. Application will access amount of time user spends on particular app. So that we can study pattern and predict level of mental pressure. If mental pressure exceeds some threshold, then user’s emergency contacts will be informed.
- Social Media Data Analysis System and Method
- Text Data Sentiment Analysis Method
- Mental health digital behavior monitoring system and method
- Sentiment analysis of context items
- Sentiment and Influence Analysis of tweets
- Mental health monitoring with multimodal sensing and machine learning: A survey
- SAD: Social Anxiety and Depression Monitoring System for College Students
- DeepMood : Forecasting Depressed Mood Based on Self-Reported Histories via Recurrent Neural Networks
- BrainNET : A Deep Learning Network for Brain Tumor Detection and Classification
- Classification using deep learning neural networks for brain tumors
- Personal Sensing: Understanding Mental Health Using Ubiquitous Sensors and Machine Learning
In Twitter Sentiment Analysis, we implemented 4 machine learning models. After analyzing their performance Logistic Regression performed the best.
In Brain Tumor Detection, we initiated with Vgg16, InceptionV3 and Resnet50 model. Among these, Vgg16 gave the best results and to enhance its accuracy we further used transfer learning method.
Furthermore, by advancing our research we worked on another 3 highly dense models such as ResNext50, FPN and Vanilla Unet. We have observed that resNext50 performed best with accuracy of 91% as it introduced a new hyperparameter cardinality which provides better fitting to model rather than by going deeper or wider. It has only one such hyper parameter to adjust.
After all implementations, accuracy of all 6 models of Brain tumor detection are as follows.
- To analyze social media account, by applying various ML algorithms, from which, Logistic regression performed best (96.288%).
- By monitoring user’s mobile activity, using which guardians, as well as doctors, can easily measure the victim’s mental health status.
- The depression check-up can update the patient’s level of depression to psychiatrist.
- In Brain tumor detection, out of 6 models ResNext50 performed best with accuracy of 91%.
- Research Gate :- https://www.researchgate.net/
- Springer :- https://www.springer.com/in
- IEEE Explore :- https://ieeexplore.ieee.org/
- Towards Data Science
- Flutter :- https://flutter.dev/
- Coursera :- https://www.coursera.org/
- Pytorch :- https://pytorch.org/
- Tensorflow :- https://www.tensorflow.org/