Predicting Review Scores
- #E-Commerce
ML model for rating customer reviews (1-5 scale) on ecommerce
- Data Analysis
- Machine Learning
- Predicting customer review scores of products
Impact
The model can be utilized to predict customer ratings based on their reviews, providing an additional data source for analysis.
Services we provided
ML model for rating customer reviews (1-5 scale) on e-commerce
Tech Stack
Python
Tensorflow
Pandas
NumPy
Challenges and Solutions
🧐 Challenges
- Procuring data to train the model.
- Developing the model for assigning scores to reviews.
💡 Solutions
- The dataset for the training was obtained from Amazon, and it includes 1 million customer reviews along with their corresponding product ratings.
- BERT with a simple regression head was trained on the data.
User flow
1. The user uploads a CSV file with customer reviews.
2. The model generates ratings for each review on a scale of 1-5.
3. The user can further process the data, such as calculating average review ratings for each product to derive scores. It helps to identify products with the best and worst scores.
4. Finally, the user can create a report about the top-performing and lowest-performing products.