Multi-level Predictive with Training Framework (MP with TF) for ranking machine learning algorithms

Machine learning algorithms differ from conventional ones in a way that they adjust themselves to perform better when exposed to surplus data. Huge volumes of data were pipelined from the grades scored by student over the last `n' years. Among the data earned from these, `k' years were populated as trained set and `(n-k)' as test sets. For the proposed Multi-level Predictive with Training Framework (MP with TF), five AI based approaches such as k-nearest neighbors, neural networks, Logistic regression, Naive Bayes and Random forest were considered to predict the target variable set. The performance of these machine learning algorithms were implemented using the actual grades scored by the students in their End Examination through a data science toolkit. Furthermore, the performances of these algorithms were ranked based on the accuracy of predicting the grades scored by the students.