Deep Learning Stock Prediction: Artificial Intelligence Expanding Applications

  2016-09-15_18-55-38  The article was written by Jacob Saphir, a Financial Analyst at I Know First.

Deep Learning Stock Prediction

“Our technology, our machines, is a part of our humanity.  We created them to extend ourselves, and that is what is unique about human beings.” Ray Kurzweil

Deep Learning stock prediction

Summary:

  • Artificial Intelligence
  • Deep Learning
  • I Know First Application
  • NVIDIA Stock Forecast and Result

Background:

What was once thought of as science fiction is now part of our everyday life.  Artificial Intelligence and deep learning are topics rarely spoken, yet it’s embedded through various applications we often overlook.  Artificial intelligence is a sub-field of computer science.  The concept of computer system to perform functions using intelligence of a human, such as: visual or audio recognition, computations, decision making, and more.  Deep learning is a subfield of machine learning.  It is composed of using artificial neural networks consisting of layers to process input data and reach its output result.  Such applications are utilized from virtual personal assistants on your phone or computer with Siri, Google Now, or Cortana to fraud detection.  And recently, it is being introduced to Amazon Go, the first of its kind to offer a shopping experience without the use of cashiers, but instead your phone and visual recognition.  By eliminating long lines and reducing labor cost, this could be the future in shopping experience.  How did we get here? How is technology continuously finding its use to be applied functionally or in search of answers?  Perhaps the biggest question we are asking ourselves is, how can we personally benefit from this?

To answer these questions, deep learning was hypothesized as early as 1965. Alexey Grigorevich Ivakhnenko theorized deep learning algorithm.  Through multiple layers of non-linear features, each data received processes through multiple layers to produce an intended result.  The diagram below shows the process:

Deep Learning

 

Imagination spanned into ideas.  Ideas became reality.  In 1989, the Bell Lab began working on utilizing deep learning to perform simple functions.  By 1993, they successfully programmed a computer to identify hand written numbers and display the numbers written, as shown in the video below:

As computers advanced with time, so did artificial intelligence’s capabilities.  On December 2nd, 2016, Steven Melendez wrote an article of a program developed in MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL).  CSAIL developed an algorithm using artificial intelligence to watch 2 million video clips and develop a 1.5 second video clip to project what the next video clip would be.  This is the neural networks using the data stored to tell the computer to predict its next event.  Although 1.5 seconds may not sound impressive to some, just the ability of artificial intelligence predicting the future result is the start of many possibilities.  So how does it work?

How It Works:

Deep learning helped spawn artificial intelligence.  Deep learning consists of artificial neural networks.  As data or input is entered into the system, the data runs through these series of virtual layers.  Each data is weighted, measured, and interpreted as its moves on to its next layer.  As Melendez’s article explains, each data taken will go through these series of layers to “collectively uncover more complex patterns in the data, expanding its understanding from pixels to basic shapes to features like stop signs and brake lights.” The data recognizes patterns and then predicts the future outcome.  All these are calculated by a series of matrix multiplications to weight the data points.  As time passes, programmers will continuously make modifications.  The larger the data and the more modifications made tweaking the weights by the programmers, the fewer mistakes the system will make in the future.

 

Deep Learning

“Deep neural networks are performing better than humans on all kinds of significant problems, like image recognition, for example,” says Chris Nicholson, CEO of San Francisco startup Skymind.  Such a technology could be advanced and applied through various aspects, such as security to monitor unusual activity, assisting surgeons during complex operations, improve reliability of self-driving cars, or predicting stock market.  A very financially rewarding application.  One who can perfect its application stands to reap immense rewards.  I Know First has been applying this technology with its predictions.

I Know First

Deep LearningI Know First, Ltd. is a financial technology company that provides daily investment forecasts based on an advanced, self-learning algorithm. The Algorithm was developed by Dr. Lipa Roitman, a scientist with over 35 years of experience in the field, and who now leads our Research & Development team to further develop and enhance the algorithm. Dr. Lipa Roitman is an R&D Chemist with a long record in computer modeling of processes, product development and process development. The concept of the current algorithm has crystallized following years of prior research into the nature of chaotic systems.

Deep Learning

Co-Founder and CEO of I Know First Ltd., Yaron Golgher has over 15 years of experience leading projects based on forecasting algorithms and predictive analytics, as well as developing quantitative trading strategies, algorithmic trading applications, and big-data solutions for hedge funds and institutional clients.

Investors are constantly in search for strategies and tools to seek consistent or high return on investment given the market’s risk.  The benefit of utilizing deep learning is its ability to process large amounts of data.  It is the challenging for any investor to process such large amounts of data while ignoring the “random noise”.  The advancement in the use of algorithms and artificial intelligence now accounts for 60-70% of “Buy” and “Sell” orders account of the US equity market volume.

The I Know First self-learning algorithm is used in quantitative trading. This form provides valuable market insight to retail and professional trader alike that is used in conjunction with traditional forms of analysis. Algorithmic traders benefit from this “second opinion” in their decision making process by verifying their own analysis or discovering new market opportunities while still maintaining complete control of their portfolio.  These algorithms analyze the structure and the trends in the market, find predictable patterns, and investors trade upon these machine-derived forecasts. This form of trading is very suitable for most investors, retail or professional.

While we cannot speak on every algorithm meant to predict the market, the I Know First market prediction system is based on artificial intelligence (AI), machine learning (ML), as well as utilizes elements of artificial neural networks and genetic algorithms. Machine learning provides an innate acumen to our comprehension of market dynamics and behavior. The algorithm has a built-in general mathematical framework that generates and verifies statistical hypotheses about stock price development. Machine learning tools such as artificial neural networks make this prediction system self-learning, and consistently determined to become more precise. This framework is used to generate initial testing models over a test sample of data. The goal of this phase is to validate the accuracy of the algorithm as well as to fine-tune the fitness function, which represents the actual goal of the algorithm expressed as a mathematical function. When the algorithm finds the global minimum of the fitness function attached to one of the models generated, it fulfills its goal.

Then a learning and prediction cycle is run with the new data included. The algorithm subsequently produces predictions for over 1,400 assets with six time horizons for each. It separates the predictable part from stochastic (random) noise and then creates a model that projects the future trajectory of the given market in the multi-dimensional space of other markets.

The I Know First algorithm returned 60.66% in 2013, beating the S&P 500 by over 30%. These results are constantly improving as the algorithm learns from its successes and failures.  The algorithms continuously beat the S&P 500 index performance throughout the years to its most recent performance in from January to November 2016 shown below:

Deep Learning

To see further proof of artificial intelligence popularity and its future in utilization, let’s look into Nvidia Corporation (NVDA).  According to a previous article posted by I Know First, NVIDIA had a bullish forecast on June 21, 2016 as the company made significant investment, such as its $2 billion research into its latest GPU interface.  GPU platforms has greatly enhanced AI capabilities.  NVIDIA is now a leader in computing platform and is now looking to become a leader in mobile platform.  These resulted in the stock to increase by over 200% YTD.  Given its success, I Know First’s algorithm is looking to achieve a 100% accuracy of its forecast released to subscribers on June 21, 2016 for the time span of 1 month, 3 months, and 1 year since the release of the forecast.  So far, they have achieved so for its 1 month and 3 months forecast with a return of nearly 100%. If the trend continues, I Know First’s algorithm will achieve its goal of 100% forecast.

Deep Learning

Conclusion:

As we can see, Deep learning has helped artificial intelligence advance in practicality in ways we have not foreseen in the past.  From 1965 models to MIT developing an algorithm to predict the future images to its use in the stock market.  Artificial Intelligence has proven itself to be a lucrative tool in stock forecasting.  As the AI processes more data, the more accurate it becomes.


premium