Our model is able to classify the emotion of users into 2 categories (positive and negative) and achieves over 80% of accuracy, which outperforms some of the classification performances presented in other research papers. Our web-application is able to handle a few users’ daily request. Although we have achieved our basic goal, there are still many work left for the future. Future work concerns improving the accuracy of classification of our model and increasing the scalability for our web-application. Due to the lack of time, the training set only contains data collected from self-assessment of our team members, in order to increase the classification accuracy, we can increase the size of training set by finding more participants to collect data from. In order to make our web-application more scalable to handle more requests, we must deploy our web server on cloud, which requires us to develop an end device that can interact with out web-application to send data to our server.
Conclusion
We developed a machine learning model that can classify the emotion of user into 2 categories: positive and negative. On top of our model, we implemented a web-application that is able to monitor, analyze and visualize the real-time emotion condition of user then provide video and advice based on the result. Besides, our web-application allows user to track the history record of emotion in order to provide more detailed analysis. During the project we have a thorough understanding of how to build a IOT web-application, especially how to divide the application into different modules to implement independently and work together to make them functional. Although we have basically achieved the goal we set in our proposal, improvement can be made by collecting more data to improve the emotion recognition ability of our model and deploying our web server on cloud to increase the scalability.