Matching and recommendation engine
Kumar, Rishu (2024)
Kumar, Rishu
2024
All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2024122738136
https://urn.fi/URN:NBN:fi:amk-2024122738136
Tiivistelmä
This purpose of this study was to create a hybrid recommendation engine for the online platform BuzziWork, which connects users with professionals in training, wellness, and tutoring services. This contrasts with product-based recommendations, because the BuzziWork engine provides specific service recommendations based on the requirements of the user, locality, and expertise of the provider.
Therefore, the aim of this research was to derive a hybrid framework that encompasses CF and CBF methodologies. CF helps to identify patterns as well as similarity in users' behavior, while the approach of CBF improves the recommendation accuracy because it focuses on the analysis of service attributes and the user preferences. This hybridization solves the cold-start problem wherein shallow or insufficient information about new users or providers hampers the quality of recommendations.
The Faker library was utilized in simulating user profiles, provider characteristics, and also the booking histories to produce synthetic datasets in order to train and test the recommendation model. The performance of the engine was measured in terms of root mean square error and mean absolute error for the amount of matching suggestions with the preferences of the users. The validation results showed that the engine was always able to make suggestions that match the preferences of the users and even in low-data scenarios.
It results suggests that AI-driven recommendation tasks produce in further user engagement and retention as well as the better targeting ability for service providers. This study shows the possibility of recommendation systems in furthering better user satisfaction and efficiency in service-oriented platforms.
Therefore, the aim of this research was to derive a hybrid framework that encompasses CF and CBF methodologies. CF helps to identify patterns as well as similarity in users' behavior, while the approach of CBF improves the recommendation accuracy because it focuses on the analysis of service attributes and the user preferences. This hybridization solves the cold-start problem wherein shallow or insufficient information about new users or providers hampers the quality of recommendations.
The Faker library was utilized in simulating user profiles, provider characteristics, and also the booking histories to produce synthetic datasets in order to train and test the recommendation model. The performance of the engine was measured in terms of root mean square error and mean absolute error for the amount of matching suggestions with the preferences of the users. The validation results showed that the engine was always able to make suggestions that match the preferences of the users and even in low-data scenarios.
It results suggests that AI-driven recommendation tasks produce in further user engagement and retention as well as the better targeting ability for service providers. This study shows the possibility of recommendation systems in furthering better user satisfaction and efficiency in service-oriented platforms.