funding-predict

This project aims to predict funding based on crunchbase data. It consists of 2 parts:

• A web app that uses ml5.js to train a neural network in the browser
• A data loader that outputs a json list of crunchbase organizations along with details of latest funding

Usage

Live Demo

The demo allows you to:

1. Choose features for prediction (expensive to consider all of them)
2. Train a simple neural network i.e. Sequential MLP with single hidden layer
3. Observe training performance, and increase training epochs if necessary
4. View most promising companies founded since Jan 2019

Note that some feature combinations may cause performance issues. Please be patient if the browser seems unresponsive. Refresh page if out of memory.

Setup Web App

The Web App was built using Vue CLI Instant Prototyping, so requires the following global packages.

npm install -g @vue/cli @vue/cli-service-global

To run locally

npm run start

To build for deployment

npm run build

Setup Data Loader

First configure database connectivity by entering credentials in .env file. No need for SSL when connecting locally.

HOSTNAME=localhost
DATABASE=postgres
USERNAME=postgres
PASSWORD=postgres
#SSL_CA=server-ca.pem
#SSL_KEY=client-key.pem
#SSL_CERT=client-cert.pem

Then run the following command to generate companies.json.

npm i && npm run data