An open API service indexing awesome lists of open source software.

https://github.com/jeffersonlicet/ml-titles-classification

💪 Public Leaderboard Top 10% - 0.89546 #MeliDataChallenge MercadoLibre Data Challenge
https://github.com/jeffersonlicet/ml-titles-classification

keras machine-learning nlp tensorflow

Last synced: about 2 months ago
JSON representation

💪 Public Leaderboard Top 10% - 0.89546 #MeliDataChallenge MercadoLibre Data Challenge

Awesome Lists containing this project

README

          

### Public Leaderboard Top 20 - 0.89546

### The challenge

The challenge was very interesting, classify articles using only its titles.

### The prize

First and second place will receive tickets to [KHIPU](https://khipu.ai/). From 3 to 5 place an Intel Movidius.

### The result

I managed to be in the top 20 with a score of 0.8954, there were more of 150 participants, the competition was hard and exciting. Of course, i learned a lot of things.

## Preprocessing

Run ml_normalization.py to generate
- categories.npy
- titles.npy
- labels_normal.npy


## Generate Dictionary
Run ml_generate_dictionary.py to generate
- dictionary.npy


## Transform tokens
Run ml_index_tokens to generate
- hashed.npy

## Training the model
Run ml_train.py to train the model using
- hashed.npy titles as x data
- labels.npy as y data
- dictionary.npy to get vocabulary


## Preprocess test data
Run ml_test_normalization.py to normalize test data and generate
- test_titles.npy


## Generate submission file
Run ml_classify.py to generate the submission file