Tailoring the Knowledge Data Discovery process to e-commerce reviews. = Tailoring the Knowledge Data Discovery process to e-commerce reviews.

This thesis mainly concerns data driven technologies and the related potentiality to bring out, from textual data, previously unknown knowledge. The goal, for the current work case, is to extract potentially useful information from Amazon products reviews. To this aim, Knowledge Data Discovery (KDD) process, applied on textual data collection, has been tailored to Amazon reviews dataset. First of all a broad description of text mining with its benefits and pitfalls has been introduced along with existing algorithm and methodologies; then, ESCAPE engine has been studied, tailored and proposed as a not-time consuming and low-computational cost solution. The proposed tool approaches and integrates all the building blocks of KDD processes such as data processing and characterization, self-Tuning exploratory data analytics, and knowledge validation and visualization. Two different approaches with self-tuning algorithms for the exploratory phase are also included. Before running the experiments, Amazon Web Service platform (AWS) case and the importance of data analysis for business choice have been discussed. A large number of experiments, applying the two approaches, have been performed on almost twenty products reviews dataset, some of them specifically built during the development of the thesis according to the work needs. Experimental results have been finally analysed, validated and visualized with several techniques in order to show, from a technical point of view, the performances of both the two ESCAPE approaches and the strategies used within them, and, from a business analysis point of view, interesting features among the different products categories comments.