Scribd
Upload
10 views1 page

2025MAI0002ADAM

SmartSiteAI is an AI-based system designed to predict restaurant ratings in North Riyadh by analyzing factors such as business type, price level, and location. The model, trained using LightGBM, achieved a RMSE of 0.3495 and a 20% Top 1 success rate in offline validation. Future work includes expanding data collection and enhancing model features for improved accuracy.

Uploaded by

squishywushy247
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
10 views1 page

2025MAI0002ADAM

SmartSiteAI is an AI-based system designed to predict restaurant ratings in North Riyadh by analyzing factors such as business type, price level, and location. The model, trained using LightGBM, achieved a RMSE of 0.3495 and a 20% Top 1 success rate in offline validation. Future work includes expanding data collection and enhancing model features for improved accuracy.

Uploaded by

squishywushy247
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 1

Alfaisal Engineering Design Expo & MAI

7th Boeing Engineering Student Competition


2025MAI0002ADAM

SMARTSITEAI: AI-BASED RESTAURANT LOCATION AND RATING PREDICTOR


Members: Eng. Abdullah Adam
Supervisor: Dr. Abd-Elhamid Taha and Dr. Nidal Nasser.

Introduction Model Training

Finding an optimal restaurant location in urban areas is challenging Model: LightGBM Regressor
due to customer behavior uncertainty. SmartSiteAI predicts RMSE: 0.3495
R² Score: 0.1444
expected restaurant ratings based on business type, price level, and
location cluster to support better business decisions in North
Riyadh.

SmartSiteAI Workflow

Collect data from Train LightGBM


Google Places API Regression Model

Evaluate with Offline Build Smart


Validation (Hit Rate) Recommendation System
Fig. 1. SmartSiteAI follows a structured workflow starting from real-world data collection, training a machine learning model, building
an AI-powered recommendation system, and finally evaluating its reliability through offline validation (Hit Rate).

Fig. 3. The dataset was collected from Google Places API and includes restaurant ratings, user engagement metrics, price levels, and

Data Collection & Feature Engineering location data. Feature importance analysis indicates that customer engagement and business type are the primary predictors of
restaurant success. Feature importance analysis was conducted to provide transparency into model decision-making, identifying
customer engagement and business type as the most influential predictors.

North Riyadh was selected due to its high density of new and diverse
restaurants, active urban development, and growing economic hubs like KAFD
Recommendation System & Offline Validation
(King Abdullah Financial District). This made it an ideal region for modeling
customer behavior and evaluating restaurant location strategies.

Fig. 4. SmartSiteAI recommends the best areas for new


restaurants based on business type and price level. The
system predicts expected ratings across different clusters
and ranks them. Offline validation demonstrates a 20%
Top 1 and 30% Top 3 success rate, confirming practical
reliability.

Conclusion & Future Work


Conclusion:
Developed SmartSiteAI to predict restaurant ratings based on business type, price level, and
location cluster.
Achieved reliable performance with RMSE = 0.3495 and R² = 0.1444.
Built a recommendation system that suggests the best areas for new restaurants based on expected
ratings.
Fig. 2. 988 restaurants collected from North Riyadh and clustered into five main areas based on geographic coordinates (latitude and Validated system reliability through offline testing, achieving a 20% Top 1 and 30% Top 3 hit rate.
longitude) using unsupervised clustering. This clustering enabled SmartSiteAI to incorporate area-specific influences into the rating Future Work:
prediction process.
Expand data collection to cover full Riyadh and other cities.
Incorporate median engagement scores to improve model stability.
This work is an expanded version of an earlier one made in collaboration with Loay Abuowaimer and Kholood Alnutaifi. Fine-tune model hyperparameters for better recommendation accuracy.
Explore additional features such as restaurant type subcategories and demographic data.

You might also like