Technical University of Crete (TUC)
Ανάπτυξη πολυκριτήριου συστήματος συστάσεων για προσωποποιημένο μάρκετινγκ
Abstract
dc:descriptionOne of the major questions for web users, is the way they should handle the huge volume of available information, to result in the selection of products that fit their needs and preferences. At the same time, corporations that sell products or services through the web are systematically trying to detect methods in order to decode users’ preference models, targeting to increase sales through better and focused product customization.Following the aforementioned questions, the research and scientific community that studies data analysis and marketing, has focused its efforts in the development of methodologies that will find proper and optimized answers to these questions. The majority of these methodologies result in the development of adaptive systems that use as input, data from the web and export suggestions for the users. These types of system are usually characterized as Recommender Systems.In the present Dissertation we present the methodology and results of the pilot use, of a new hybrid recommender system that is based in the use of Multicriteria Analysis, Sentiment Analysis and Filtering methods. The methodology results in 4 different types of recommendations.Within the methodology we manage to define the preference models of system users, which are later correlated with relevant “customer profiles” who have already chosen specific products/services.Thus, we result in personalized product recommendations for the system user, that fit her preferences. Additionally, she is given the opportunity to filter the product alternatives by choosing among a number of flat criteria, the ones that she considers as prerequisite characteristics of the final recommendations. The use of sentiment threshold results from the sentiment analysis on customers online reviews and acts as quality guarantee.The input data of the system are real online users ratings and reviews which are mined from the web using crawlers.System development phases are based in CRISP-DM Methodology. The evaluation of the recommendation system is implemented by measuring the prediction accuracy in real users experiments.The case study occurred hotels in Chania region. The most important characteristics of the specific research are the following:a.Use of big volume of real data (mined from the web), in contrast with most of the studies that use ready to use test sets.b.Use of all the available mined information in order to result to the recommendation. More specifically, we use Customer Ratings data to implement Customer Satisfaction Analysis, Customer Reviews Data for Sentiment Analysis, Static product data for the flat criteria filtering.c.Use of sentiment analysis results as recommendation quality thresholds.d.Successful development of a new approach for preference modeling, using WAP.e.Minimum data input from the system user.f.Minimum time to produce recommendation, due to the system architecture.g.Avoidance of use of personal data for the recommendation process.h.Avoidance of Cold Start problem.i.Satisfactory experimental results according to the error metrics that were used for the evaluation of the system.The main metrics that were used to evaluate the quality of recommendations were the following: Mean Square Error (MSE), Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE).The dissertation sums up with suggestions for future research and extension of the new recommendation system.
Degree
thesis:*- Grantor dc:publisher
- Technical University of Crete (TUC)
- Year dc:date
- 2018
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Zisos, Ioannis
- Ζήσος, Ιωάννης
Subjects
dc:subject × 20- Συστήματα συστάσεων
- Πολυκριτήρια ανάλυση αποφάσεων
- Ανάλυση συναισθήματος
- Τουριστικό μάρκετινγκ
- Έμπειρα συστήματα στο μάρκετινγκ
- Πολυκριτήρια Συστήματα Συστάσεων
- Recommender systems
- Multicriteria decision analysis
- Sentiment analysis
- Touristic marketing
- Hybrid intelligent systems
- Multicriteria Recommender Systems
- Φυσικές Επιστήμες
- Επιστήμη Ηλεκτρονικών Υπολογιστών και Πληροφορική
- Επιστήμες Μηχανικού και Τεχνολογία
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Natural Sciences
- Computer and Information Sciences
- Engineering and Technology
- Electrical Engineering, Electronic Engineering, Information Engineering
Rights
dc:rights- Statement dc:rights
-
- BY_NC
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/44017