Machine Learning and Human Activity Recognition: Building a Classifier for Wearable Accelerometers’ Data
#DataScience
#RStats
#MachineLearning
#HumanActivityRecognition
#Wearable
Implement a machine learning model to predict if body postures and movements when a unilateral dumbbell biceps curl repetitions are being correctly realized.
Data was collected from accelerometers' used in devices such as Jawbone Up, Nike FuelBand, and Fitbit.
The CRISP-DM (cross-industry process for data mining) methodology provides a structured approach to planning a data mining task and was applied to this project.
In order to classify if each of the repetitions were made according to the specification (Class A), throwing the elbows to the front (Class B), lifting the dumbbell only halfway (Class C), lowering the dumbbell only halfway (Class D) and throwing the hips to the front (Class E), Random forest Machine Learning algorithm was applied to the 52 variables collected from the sensors.
Random forest has demonstrated to be a solid choice for this classification problem. This relatively new machine learning strategy (it came out of Bell Labs in the 90s) belongs to a larger class of machine learning algorithms called ensemble methods which involves the combination of several models to solve a single prediction problem.
Model robustness for making predictions on unseen data was evaluated using cross validation and classification accuracy (it is the number of correct predictions made divided by the total number of predictions made, multiplied by 100 to turn it into a percentage).
The final model achieved 99.3% accuracy on its classification task
Files:
har.md
: a markdown script with the project.har.Rmd
: a R markdown script for reproducible research .har.html
: R markdown compiled .solution.csv
: Labeled test data.
Jose Ramón Cajide Data Scientist at El Arte de Medir
For more information about the Weight Lifting Exercises Dataset and the project visit: http://groupware.les.inf.puc-rio.br/har