Project Overview

The purpose of this project is to provide an all in one machine learning portfolio management package for R.

The package will.

  1. Provide a framework for downloading and creating datasets of important data
  2. Generate an optimal portfolio mix of assets
  3. Generate an allocation recommendation per day

Built into the software will be ways to:

  1. Manage the available assets to mix in
  2. Clean and normalize the data sets
  3. Generate forecasting models
  4. Generate Global search machine learning models for portfolio mix
  5. Generate machine learning Neural Net models for allocation

Process Flows Needed

Geared towards a Trading platform business/system

Sales:

Investor Aquisition Process Flow (Sales Team) TBD

Operations:

Minimal Viable Product Flow (Product Team)

Trading System Integration Process Flow (Systems that integrate with and execute trades with the exchange) TBD

Data Gathering and Cleaning Process Flow (Non exchange related data gathering) TBD

How to troubleshoot any Artificial Intelligence or Machine Learning system ever. No exceptions, throw out the black box. Stop sounding confused and scaring people.

This is not a tutorial it is commentary and maybe a little informative....So… In summary.

  • Statistical models are used to generate an analysis and decision making framework when we don’t have complete understanding of how a thing works

  • Artificial Intelligence is just an automated version of statistical model building

  • Tune and test your models and get a confidence level

  • If it was perfect data and a perfect model, you wouldn’t need statistics or AI...GET OVER IT OR QUIT SCIENCING

Clickbait on an MIT article? What the hell.

Basic Structure of the system

Process streams

User Interface to pick stock portfolio

  • Drupal - web component fronted by Drupal that allows the picking of stocks, display of charts, and status updates from the model building, Daily allocation and King of the hill testing

  • The interface pulls data from the R/C++ backed

  • This is what the users interact with, and the rest of the streams are not bound by it. All of the other streams and Dependant upon each other, but can work completely without the web interface

Lots of progress!

Wow have I added a lot of functionality to this stuff since november.....  let me know if you guys play with it... I'm having fun...

Neural Net Creatures building stock portfolios

Have the Neural net framework working...

Have the Genetic Algorithm Wrapper working...

Ran this over 5 generations only

50 stocks set as the available features

each generation only had 50 Neural nets in them too.

Random portfolio generation from the 50 and evaluated after the nets were trained.

Still very simple free data sets and a very small set of available stocks.

Lighting Talk at Data Philly

Did a meetup at Data Philly on what I have so far.  Made some good contacts and got some good feedback.  Generally positive with some cool suggestions on how to approach the performance function that my little creature nets are judged on.  Different investment styles will have to be plugged in as necessary.

Fast Compressed Neural Network For R Evaluation

DRAFT - Feeback is welcome

My findings in this research reinforce what was found here:

http://people.missouristate.edu/randallsexton/sabp.pdf

I've done a general analysis of the FCNN4R package for R. 

The source for it is located here if you are interrested in playing with the code.

https://github.com/keithaumiller/unicorninvesting

SA BP and SGD algorithm performance

Peco Energy Utilization clustering analysis utilizing R

General Steps and PECO 

Goal of this process is a basic clustering analysis to identify a customer via a Profile or Cluster identifier to easily be able to predict their utilization practices/forecast what their utilization will be.

I had to do this without Tableau at the time so please forgive the excell...

General process Flow for creation/setup/training/feedback loop

Pages

Subscribe to Unicorn Investing RSS