Inferring the Causal Relationships in Student Placements’ Performance using Causal Machine Learning

Abstract

This research proposes using causality models to analyse and infer student placement data. It demonstrates the distinctions between applications of Causal Machine Learning and Machine Learning for resolving different education-related processes. Association does not equal Causation. In traditional machine learning, the focus is often on predicting outcomes or patterns based on input data. However, causal machine learning goes beyond prediction by aiming to uncover cause-and-effect relationships between variables. The review of causal inference in the presence of massive data sets is a rich and expanding field of contemporary research. The goal of causal inference is to understand how changes in one variable affect another, and to identify the underlying mechanisms that lead to certain outcomes. The causal Inferencing which is the key concept for causal machine learning can be implemented using the DAG (Directed Acyclic graph). Through this paper we aim to provide some useful insights using 3 causal discovery tools (PC, GES, LiNGAM) to produce the DAGs. We proposed a novel 3D framework (Data correlation, Discovery tool using Causal ML, Domain knowledge) which combines the merits of both manual and causal discovery tools. The causal graph obtained is checked for falsification i.e. the correctness of the graph. The obtained graph needs to be informative and significance level (p-value < 0.05) so that the DAG would be accepted. Thus, a final Causal Model is formed that represents relationships between the variables to understand and predict the effects of interventions or changes in the system.