This week, I personally made a great deal of progress on the machine learning model in the backend of our web application. I began by creating some mock training data by randomizing the day, hour, minute, and wait time. I made some time periods “busier” than others by having the wait time randomized to be higher during those times. I then tried randomizing the training data and training the model a few times in order to see how accurate the neural network is with it. Although I made a lot of progress getting the loss function down, I need to make the data a bit less random in order to have it work better and make the mean squared error lower. Below is some sample output from model training and predictions for testing.
In addition, I made progress setting up the input form on the web application. I successfully implemented validation for the form to ensure that the inputs make sense (exit time comes after entry time, the time and day aren’t in the future, etc.). Below is some of my form validation code.
My progress is on schedule so far. In the next week, I hope to get the neural network accuracy up as well as begin working with Dina and Sam on the connection between the RPi and the web application. Once that component is working, it will be much more simple to connect Sam’s circuit with the RFID reader to the web app later on.