ML Specialized Project

CloudMile project for google ml-specialist

DNNRegressor/DNNLinearCombinedRegressor on Rossmann data

Rossmann data is a collection of time-series data for the prediction of sales in the next 6 months for each store. The greatest challenge is whether the model can accurately obtain the mean sales for each store, as well as the periodic and seasonal information.

• From kaggle Rossmann
• See the rossman.ipynb for details

Custom tf.estimator.Estimator for KKBOX Music Recommendation Engine

KKBOX data is the data for a music recommendation challenge. Through personal records, we attempt to predict the scoring for songs as well as the click-through rate (CTR).

• For this personalized music recommendation, since the target column is binary, we tackle this task as a classification problem, i.e. this is a binary classification problem.
• From kaggle WSDM - KKBox's Music Recommendation Challenge
• See the kkbox.ipynb for details

Train Data and Model Check Points on GCP

• Rossmann
  • data: gs://ml-specialized/rossmann/data

    │  store.csv
    │  store_states.csv
    │  test.csv
    └─train.csv
    

  • model: gs://ml-specialized/rossmann/models

    │  saved_model.pb
    └─ variables
            variables.data-00000-of-00001
            variables.index
    

• KKBOX
  • data: gs://ml-specialized/kkbox/data

    │  members.csv
    │  songs.csv
    │  song_extra_info.csv
    │  test.csv
    └─ train.csv
    

  • model: gs://ml-specialized/kkbox/model

    │  saved_model.pb
    └─ variables
            variables.data-00000-of-00001
            variables.index
    

The project can be executed with jupyter notebook through the work flow after data is stored under respective directories. You will also need a GCP account to deploy model on Google Cloud Platform.