Geospatial Insights for Retail Recommendation Using Similarity Measures

Area of Science:

• Geographic Information Systems (GIS)
• Retail Analytics
• Machine Learning

Background:

• Recommending retail businesses based on location is complex, involving demographics, trade area characteristics, sales, traffic, and environmental factors.
• Manual recommendation processes are labor-intensive and prone to inconsistency due to subjective expert opinions.
• Challenges arise when sales data is unavailable for making informed recommendations.

Purpose of the Study:

• To develop and evaluate a machine learning approach using similarity measures for retail business recommendations.
• To address the challenges of feature set preparation for diverse retail types and selecting optimal similarity measures.
• To assess the accuracy and potential biases of proposed recommendation approaches across different retail categories.

Main Methods:

• Utilized a machine learning approach centered on similarity measures for retail business recommendation.
• Developed a common feature set incorporating points of interest, population, property, job type, and education level data.
• Conducted empirical studies to evaluate recommendation accuracy and category bias.

Main Results:

• Proposed similarity-based techniques achieved recommendation accuracy exceeding 70%.
• Higher accuracy was observed when recommendations were made within a cluster of similar retail businesses.
• Empirical studies indicated no significant bias toward specific retail categories.

Conclusions:

• Machine learning, particularly similarity measures, offers a viable solution for retail business recommendation, especially when sales data is absent.
• The developed feature set and similarity approaches provide a robust framework for location-based retail recommendations.
• The findings support the effectiveness of data-driven, similarity-based methods in enhancing the accuracy and consistency of retail recommendations.