| dc.description.abstract | Our thesis addresses changes in customers’ purchasing behavior subject to the nature of limited
usage period of a product or service. We propose a novel time-weight Recommendation system
with Associaton Rules. Our time function develops a joint probability considering customer
purchasing tendency over time and data recency. For Association Rules Mining, we adopt both
Apriori and FP-growth algorithms. Depending on what step to apply the time-weight function,
we propose two different approaches, called Pre-ARM and Post-ARM. We also benchmark our
system with a Non-weighted association-rules-based system to elaborate the preeminence of our
proposed system. Through experiment data, we prove that our proposed system outperforms the
Non-weighted association-rule-based system by 30% in the Post-ARM Recommendation system.
Through analysis, we noticed that while our model helps with economic and social issues, it also
has effects on environmental issues. As a result, a balance between economic and environmental
objectives is necessary to account for both the advantages and disadvantages of our model. | en_US |
