This is a basic machine learning project based on Multivariate Linear Regression based on the cost and gradient formula.
The dataset we used is a text file having attributes separated by commas.
======= This machine learning project gives prediction of how much a customer spends Yearly using the attributes as follows: 1) Avg. Session Length 2) Time on App 3) Time on Website 4) Length of Membership
Here we implement the GRADIENT DESCENT using the help of Python frameworks namely:
1.Numpy
2.Pandas
3.Sklearn (For testing purpose only to split the testing and training data using model_selection attribute)
4.Matplotlib (To check the cost function graph with each iteration and to check the linearity of each attribute
with the expected data)
We use the pickle module of python of to save our Revised Theta obtained using the gradient descent locally to file with extension .pickle which can be directly used thereafter without training the model again.
The Learning rate we chose was alpha = 0.0003 You code use anything lesser than 0.1 so as for the model to work properly
The iterations we decided to be was 100,00,000. You can also chose around 10,000 but we wanted the accuracy to be really high. If time constraint for compiling is not a issue then keeping iterations as high as possible is preferred to keep the model from predicting accurately.