Skip to content

codenation-dev/squad-3-ad-data-science

Repository files navigation

[squad-3-ad-data-science]

This model is a recomendation system that analyze a market to recommend leads to users with a previous list of clients that are inside the market.

Stakeholders

Role Responsibility Full name e-mail
Data Scientist Author [Lincoln Vinicius Schreiber] [[email protected]]
Data Scientist Author [Murilo Menezes Mendonça] [[email protected]]
Data Scientist Author [Nathan dos Santos Nunes] [[email protected]]
Data Scientist Author [Thiago Sant' Helena] [[email protected]]

Usage

Usage is standardized across models. There are two main things you need to know, the development workflow and the Makefile commands.

Both are made super simple to work with Git and Docker while versioning experiments and workspace.

All you'll need to have setup is Docker and Git, which you probably already have. If you don't, feel free to ask for help.

Makefile commands can be accessed using make help.

Make sure that docker is installed.

Clone the project from the analytics Models repo.

git clone https://github.com/<@github_username>/squad-3-ad-data-science.git
cd squad-3-ad-data-science
mkdir workspace/data
mkdir workspace/models

Be sure to configure train and test data on squad_3_ad_data_science/config.py, placing files on workspace/data folder.

To train and generate test metadata, run

make run

Performance metadata available on workspace/performance.json

To get recomendations, run

make predict INPUT='<path_to_input_file_with_column_of_ids>' PARAMS='--k 10'

Param k stands for the number of recomendations. Default is 10.

Final Report (to be filled once the project is done)

Model Frequency

make run takes ~5 min (assuming installed libraries)

make predict takes less than 1min

Model updating

More filtering and feature tunning can be added to feature_engineering.py, more validation functions to generate extra metadata for model validation on validation.py and more recomendation functions on recomendations.py.

main.py must be updated to apply new methods.

Keep logs with loguru.

Maintenance

To deploy as web application, use functions implemented on module to make predicts with main.predict(input_file, **kwargs).

Minimum viable product

--

Early adopters

--

Documentation

Folder structure

Explain you folder strucure

  • docs: contains documentation of the project
  • analysis: contains notebooks of data and modeling experimentation.
  • tests: contains files used for unit tests.
  • squad_3_ad_data_science: main Python package with source of the model.

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 3

  •  
  •  
  •