Monitoring of Daily Activities Using Machine Learning

The numbers of fitness bands and other IoT devices such as sleep trackers etc. have risen exponentially. With the amount of data now available through the means of these devices about people from all walks of life has risen greatly too. All of the daily activities of a person amount to something so there needs to be a pattern to the amount of data collected by the means of these various devices such as a sleep tracker and a fitness band.

Currently, there are few applications which assess the data for the user. A lot of it has to be done by the user manually. We are making an application which will monitor this daily activity data through these devices and assess it and find patterns in it using k-means clustering in unsupervised learning. The assessed data will be collected in a database and store it in the cloud and use it as training data and tests will be run on it to find patterns. We further aim to predict user actions and health problems through the data collected. Such an application and assessed data can be useful to various institutions, fitness companies etc.

The main aim of this project is to develop an application which uses the collected data and assesses it to find patterns in it. Mental stress is one of the growing problems of the present society. The number of people experiencing mental stress is increasing day by day. Stress is a response of our body to prepare itself to face difficult situations. When a person goes under stress, his nervous system responds by releasing stress hormones. These hormones make our body ready for emergency actions. In certain situation, it becomes dangerous and can put a person in serious mental disorder. Long-term effects of stress can be chronic. Chronic effect of the stress causes health problems like hypertension, cardiovascular diseases, and memory problems. The sense of loneliness and hopelessness may lead people to suicide. People may be less likely to notice whether they are under high stress or may be generally less sensitive to stress. Stress detection technology could help people better understand and relieve stress by increasing their awareness of the heightened level of stress that would otherwise go undetected. For this objective, we have designed a smart band device in order to detect different conductance levels of the skin and predict whether the person is under stress or not. But skin conductance alone cannot accurately predict the stress level in everyday activities. Physiological responses caused by stress can also be provoked by physical activities like running, lacking sleep etc. In order to accurately measure the stress level, classification should be made. The fit band will be capable of detecting stress by analyzing different parameters in accordance with skin conductance like activities tracking, sleep quality etc. The collected data is then transmitted to user’s smartphone via Bluetooth and upload to the web from where it is accessed to find patterns to further ease the user experience.

The main aim of this project is to develop an application which uses the collected data and assesses it to find patterns in it. This can be done by collecting a large sample of data and using it as training data by using unsupervised learning methods such as k-means clustering to find patterns it. k-means is one of the simplest algorithms that solve the well-known clustering problem. The procedure follows a simple and easy way to classify a given data set through a certain number of clusters (assume k clusters) fixed apriori. The main idea is to define k centers, one for each cluster. These centers should be placed in a very cunning way because location causes a different result. So, the better choice is to place them as much as possible far away from each other. The next step is to take each point belonging to a given data set and associate it to the nearest center. When no point is pending, the first step is completed and an early group age is done. At this point, we need to re-calculate k new centroids as barycenter of the clusters resulting from the previous step. After we have these knew centroids, a new binding has to be done between the same data set points and the nearest new center. A loop has been generated. As a result of this loop, we may notice that the k centers change their location step by step until no more changes are done or in other words, centers do not move anymore. We are using k-means clustering methodology because it is fast, robust and easier to understand.

A basic application would require a large sample of data, collected over long periods, and with a vast variety of users, ranging from different age groups, and different genders, and different height and weight. Once we have collected this sample data, we can use k clustering to find patterns in it. Unless this data is large, k clustering might not provide a very accurate result. The learning algorithm requires a prior specification of the number of cluster centers. Randomly choosing of the cluster center cannot lead us to the fruitful result. These are some problems associated with the algorithm.

Once this data is collected using the smartband and stored in the application using IoT, the algorithm is applied, we can find patterns, which will show that all the input data, i.e. the number of steps walked, the amount of sleep, the calories burned, and the heart rate will have a relation of some sort between them. This relation will be found out by using the k means learning algorithm. All this information and data and results will be stored and the user can view them anytime. According to the results, the user can change or alter his input data, so that he can gain favorable results