ML Paper submission for UWE

Predicting product cancellations for sales retention

Running the notebook

Please install https://www.anaconda.com/download/ and then open the notebooks in Juypter. I've split it into to notebooks:

• Investigation and feature selection - does what it says
• Pipeline and prediction cross validation - The actual pipeline and ML process, this one has a section which will take a long time to run (Test classifiers - THIS IS SLOW!)

Changes to get the essay into a form that can be used for a research paper submission - Notes

Azure is OK.

• Don't forget to describe data (a sample of data provided, say a row, should be described in the paper).

• ANN, do you mean classic neural networks or those classified for deep learning (I'll help in both, my guess is that we might go for a tensorflow implementation for a deep learning one — and I will do that) — give me a week time.

• Cross-validation (for RF you don't need as randomization is per se a cross-validation approach), for the rest yes, sklearn provides the suitable libraries as in your link.

• Frame the literature review into the kind of business you are dealing with, and its impact The - impact of sale retention and the related impact of cancelation on businesses (an example?)

• We need to relate to other works done on this dataset or similar ones — do we have?

• Expanding the algorithmic solution to include, for instance, Random Forests (and others, say, SVM, NBC) and I will help in Neural Networks.

• Remove the justification and literature about ROC and evaluation methods — we’ll just use them and analyze the resulted curves.

• Conclusions should relate to the obtained outcomes, mainly. Let’s not go far with coursework-specific items in all the document.

• Cross validation is required for evaluation of algorithms — I will help in this

We need to relate to other works done on this dataset or similar ones — do we have?

A brief look at existing papers: https://ieeexplore.ieee.org/document/8014605/ This is interesting: https://ieeexplore.ieee.org/document/4666835/ I’ll read these tomorrow and see if they apply to this problem.

Conclusions should relate to the obtained outcomes, mainly. Let’s not go far with coursework-specific items in all the document.

OK I’ve setup the document on here: https://v2.overleaf.com/3526875447bghttgtbhzvy with the correct LaTeX formatting, I’ll move the essay in and edit it there – you should be able to edit directly or add comments. I’ll work on this tomorrow evening and ping you with any updates.