Human Activity Recognition

Overview

This project implements machine learning classification of accelerometers data on the belt, forearm, arm, and dumbbell of 6 participants to predict the manner in which people perform the exercise.

Background

Human Activity Recognition (HAR) has been recognized as a key research area and is gaining attention by the computing research community, especially for the development of context-aware systems. There are many potential applications for HAR, like: elderly monitoring, life log systems for monitoring energy expenditure and for supporting weight-loss programs, and digital assistants for weight lifting exercises.

Data Exploration

I use pandas to deal with data in the .csv file. There are total 160 categories (features) but some are useless, such as raw_timestamp_part_1, new_window, and so on. Hence, I drop those features and use the remaining 52 features to train the model. Also, since there are no labels in the testing data, I split the training data into training set (75%) and validation set (25%), and perform cross validation (5-folds) to compute the accuracy. The details of prepocessing is available at utils.py.

Prediction Modeling

At first, I try to use the SVM model with RBF kernel but it does not work perfectly. Then, I turn to Random Forest Classifier and it has much better performance. As for the hyperparameters, I use grid search in order to achieve the best performance.

• SVM
  • kernel: ‘rbf’
  • C: 5
  • gamma: ‘scale’
  • Tuning:
    • C: [1, 2, 5, 10]
    • gamma: [‘scale’, ‘auto’]
• Random Forest Classifier
  • n_estimator: 500
  • max_depth: 500
  • class_weight: ‘balanced’
  • Tuning 1:
    • n_estimators: [10, 100, 200, 500]
    • max_depth: [200, 500, None]
    • criterion: [‘gini’, ‘entropy’]
    • class_weight: [‘balanced’, None]
      
  • Tuning 2:
    • n_estimators: [500, 1000, 2000]
    • max_depth: [250, 500, 750, 1000, None]
    • class_weight: [‘balanced’, None]
      
  • Note: It is reasonable that the more n_estimatros and max_depth, the higher the accuracy.

Note: Both have 25% validation set and do 5-folds cross validation.

Model Application

The following are the results of 2 models. Please read the jupyter notebook (here) for more details.

• SVM
  • accuracy: 93.31%
  • prediction: [‘B’, ‘A’, ‘B’, ‘A’, ‘A’, ‘E’, ‘D’, ‘B’, ‘A’, ‘A’, ‘B’, ‘C’, ‘B’, ‘A’, ‘E’, ‘E’, ‘A’, ‘B’, ‘B’, ‘B’]
• Random Forest Classifier
  • accuracy: 99.69%
  • prediction: [‘B’, ‘A’, ‘B’, ‘A’, ‘A’, ‘E’, ‘D’, ‘B’, ‘A’, ‘A’, ‘B’, ‘C’, ‘B’, ‘A’, ‘E’, ‘E’, ‘A’, ‘B’, ‘B’, ‘B’]

As we can see, both models predict the same result and I believe they are good enough for this problem. The performance of Random Forest Classifier is better than SVM and it classifies almost correctly on the validation set. Besides, Random Forest Classifier takes less time to train the model so it is more effiecient.

Testing Environment

• sysname: Linux
• release: 4.15.0-45-generic
• machine: x86_64
• python: 3.6.7 and 3.7.1

Required package

• scikit-learn