HEALTH PREDICTION BASED ON ACTIVITY PATTERNS BY USING MACHINE LEARNING TECHNIQUES

Abstract

Nowadays Healthcare industry has become a big business. Science is advance and is able to provide treatment of any disease using advance machines used in pathology center. But the treatment would be more helpful if the disease of patient will be known in advance stage. There are lots of online system available, which can help patients to know there disease based on taking few input from patients. This system supports an end user and online consultation. These input are related to the information which they feel while having problem. The system will treat this input/information as a symptom of the disease. The system will match this symptom with the previously available data on the database, which they collected from the doctors/patient from the previous records. Finally the system will predict the disease of the patient. But there should be one such system available which will take care health of the person. Which will help them to make them healthier by giving them feedback base on the daily activity they perform. If such system will be available then the chances of having disease will be reduces. Since peoples are very much aware of his/her health. To become healthier they do regular exercise and consult with doctor for yearly medical checkup. But they don’t know when their regular activity will spoil their health. In this thesis, we have presented a model to predict health based on person’s lifestyle, which is formed by using daily activities performed by the person. To implement this health prediction system, we have recognize human activities by taking data from sensor-rich smartphones. After this we have used these human activities to create user life document, which represents user’s life style. By measuring lifestyles of users is we can predict the health and suggest them to change in the lifestyle. First, we collected sensor data of user using mobile application and then we performed activity recognition. After finding activities of user we have created a life document of user, from which the algorithm Latent Dirichlet Allocation (LDA) [1] is used to select the life style. We further propose a similar metric algorithm in order to compute the health of users to measure the life style data available in training data, and measure the impact the life style of users by graph. When user want to know their health, then our algorithm will return list sorted by activity score, from which user can know about his health and take prevention to not having any Page 4 disease. We have implemented this system for Android based smartphone and its performance has been evaluated on with large-scale experimental data. Finally, the prediction results manifest that the prediction properly reflecting the preferences of activity perform by the user in day, week or month. This approach exploits gradient boosting algorithm, Auto-regression Model, Signal Magnitude Area (SMA), tilt angle, standard deviation mean & median.