Understanding Text Features and time importance of a machine learning model on plumbers reviews

Abstract

Machine learning has a lot of potential for improving products, processes, and research. But the model usually does not explain its predictions which is a barrier to the adoption of machine learning. The model must also explain how it came to the prediction because a correct prediction only partially solves your original problem. This paper use reviews scraped from thesite werkspot.nl, a Dutch online market platform for small construction work, to analyse the sentiment reflected in the written review, and to predict the rated amount of stars. This paper aims to understand the importances of the word inside the model and trying to see the differences after including time context inside the model. To understand the feature importances, I use Shapley Analytics value to measure the feature importances inside the model and Mean Absolute Error and Root Mean Squared Error to measure the model performance. The main finding is the word importances before and after including the time context change some importances of some words inside the model and improve the model performance