Assignment 3

Due date: Thursday, May 21st, 2026 at midnight. 

Consider the Human Activities and Postural Transitions Data Set The data set consists of motion data recorded (at 50Hz) by smart phones attached to the waists of people undergoing 12 different activities. I've reduced the data time series down to 6 features; this version of the data can be found here.  The last column of this file is the target category.  The goal of this assignment is to use the timeseries data to categorize the activity being performed by the subject.

Create a Python script, called "hapt_lstm.py". Your script should:

• read in the data set and separate out the time series from the targets,
• break the timeseries data into chunks of whatever length seems appropriate (note that the chunks must all be of the same motion category),
• split the data into training and testing data sets,
• build an LSTM neural network, using Keras, to predict the motion category, given a chunk of the timeseries,
• train the network on the training data, and print out the final training accuracy,
• evaluate the network on the test data, and print out the test accuracy.

Your script will be tested from the Linux command line, thus:

$ python hapt_lstm.py
Processing data.
Building network.
Training network.
The training result is [0.0144, 0.9950]
The test result is [0.0170, 0.9938]
$

You do not need to use the full data set (it can be quite large, depending on how you set things up), but you should use enough of the data such that overfitting is minimized.

Submit your hapt_lstm.py file. 

Assignments will be graded on a 10 point basis.
Due date is May 21st 2026 (midnight), with 0.5 penalty point per day off for late submission until the cut-off date of May 28th, at 11:00am.