AI-nance: Stocks Prediction

Final thesis diploma for the Software Engineering Degree of the Universidad Politecnica de Madrid.

Overview

The goal is to build an entire infrastructure using a micro-services pattern, focusing on automation and deployment, to serve NASDAQ stock prices predictions.

Architecture

How to run

Tensorflow Server

You need to export the environment variable $MODELPATH in order for Tensorflow Server to be able to find the prediction model. Tensorflow Server will automatically pick the models from /predictionModule/mlp_model, save your models there to be able to access them.

export MODELPATH="/repo-location/predictionModule"

Docker Compose

After installing Docker and Docker compose, you can run the compose command in the same location as the docker-compose.yml file.

docker-compose up

Once all the containers are up and running you can request price predictions by specifiying the desired company ticker.

localhost:8000/predition/<str>:ticker