Comparing ML Methods for Facebook Post Classification

During my third year of university I carried out a collection of Machine Learning projects as both an individual and as part of a team for a module 'Data Analytics and Machine Learning'. The goal of these projects was to strengthen my ability to build machine learning models in Python using libraries such as ScikitLearn and Pandas, as well as analyse these models, and hone the skills necessary for critiquing and improving them, by understanding how to interpret metrics such as Accuracy, Precision and Recall, and using these results to iteratively tweak the parameters of the model.

My individual project for this module was a comparison of multiple different machine learning methods on a set dataset, with the aim of identifying the most effective methods and quantifying their success for a classification task.

I used a dataset of Facebook posts tied to accounts using Facebook as a platform for ecommerce sales, and set out to investigate whether using only the engagement metrics from a collection of their posts, if we could successfully categorise the Facebook activity into either a live stream (video) or a simple Image (photo) post.

I was generally successful in this classification attempt. The full report outlines just how successful this classification was for a range of machine learning techniques, and attempts to identify the most effective for a classification problem of this nature.

You can view the full report by clicking here.