Human Activity Recognition

The dataset we used contains the recordings of the experiment Human Activity Recognition with Smartphones, which has been carried with 30 participants between 19-48 years old doing their daily activities. The daily activities performed by the participants are WALKING, WALKING_UPSTAIRS, WALKING_DOWNSTAIRS, SITTING, STANDING, LAYING. During the recordings, participants wear embedded accelerometer and gyroscope of Samsung Galaxy S II on the waist to capture 3-axial linear acceleration and 3-axial angular velocity at a constant rate of 50Hz.The recordings of the participants are separated into 2 groups as 70% for training data and 30% for test data. The training data consists of 7353 rows and 563 columns. The test data consists of 2948 rows and 563 columns. Each row contains a record of a participant for different activities. 561 columns are features of the records, 1 column is the person identifier and the last column is the activity label. Features and labels are separated into two different tables. In order to increase the accuracy and decrease the bias and the variance of the model, we combined train and test data, and split them again by using k-fold cross validation, k is 10 in our case.