multi_label_text_classification

Multi-label text classification

In this project, we want to classify input samples of the Ronash dataset into 20 categories via its title, vendor, and tags.

In this notebook, I present the updated version of the multi-label text classification model created by Farakh Karimi. The original model achieved an accuracy of 81.21% and a loss of 0.8373. In the updated version, the following changes have been made to improve the model performance:

Exploratory Data Analysis (EDA):

• Distribution of Products by Vendors
• Pie Chart of Product Distribution by Category
• Histogram of Number of Tags per Product
• Bar Plot of the Top 10 Most Frequent Tags

Class Balancing:

• Data analysis by grouping based on the 'category' column and counting the number of products in each category (category_counts = df['category'].value_counts()).
• Correcting class imbalance using upsampling.

Improving Text Transformation Layer Parameters:

• Re-defining max_features and sequence_length parameters:
  • max_features = 19000 (increased from 10,000 to 19,000)
  • sequence_length = 320 (reduced from 350 to 320)

Model Performance Evaluation:

• Improved accuracy and reduced loss:
  • Loss: 0.1043
  • Accuracy: 0.9786

Performance Analysis:

• Plotting the confusion matrix and classification report for detailed analysis of the model's performance.

This update demonstrates significant improvements in the performance and stability of the multi-label text classification model.