aws-samples / asset-prediction-sagemaker-pipelines-example Goto Github PK

View Code? Open in Web Editor NEW

Asset Prediction Sagemaker Pipelines Example

License: MIT No Attribution

JavaScript 24.58% Shell 0.70% Python 5.25% TypeScript 68.73% HTML 0.12% CSS 0.12% Raku 0.17% Perl 0.05% Jupyter Notebook 0.26%

asset-prediction-sagemaker-pipelines-example's Introduction

Asset Prediction Sagemaker Pipelines Example

⚠️ Disclaimer: This code is provided AS IS. Check out the LICENSE file.

⭐ Introduction

This Asset Example provides a prototype-level end-to-end ML Prediction SageMaker Pipeline for customers who want to use Amazon Sagemaker to predict (financial) timeseries data.

This example leverage the prediction model introduced in the blog post Enhancing trading strategies through cloud services and machine learning, but can be customised to any other market or any other timeseries type of data.

This example pushes the blog's code one step further by automating the whole deployment pipeline, leveraging AWS SageMaker:

Automated listing of the datasets from the S3 origin buckets
Automated Feature engineering on the data selected
Automated model training with possible customization of the model's hyper-parameters
Visualization of the status of the pipeline, its tasks and of the model inference results

This project also showcases the implementation of the following major components:

Cloud deployment of (serverless) resources as Infrastructure as Code (IaC)
Web Application, to interact with the pipeline and visualize the model's result and to interact with the platform
Example scripts to gather data and load into the system

Getting Started

Folder structure

.
├── apps
    ├── asset-import (scripts to download and import assets)
    ├── infra (cdk application to deploy cloud resources)
    ├── scripts (scripts for local development)
    └── website (web app)
├── docs (detailed documentation)
└── packages
    ├── @config (project config related packages)
    └── @infra (lerna packages for infra deloyment)