Integrating next generation, Big Data, AI-driven machine learning systems

Contact Us

30 Wall St, 8th Floor
New York, NY 10005

info@predictivetradingsystems.com
(212)365-4525

WHAT WE DO

Predictive Technology Systems is a leader in the transition from human-guided financial trading to fully-automated systems. The convergence of machine learning software (ML), digital language processing (DLP), sentiment analysis, and massive cloud storage technology enables today’s asset managers to use AI-guided programs to analyze years of historical data along with real time current events, allowing data rich analysis completely free from human emotion and biases.

Core Technologies of our Finance-as-a-Software-Service (FaaS) Platform

MACHINE LEARNING

At the heart of AI-driven trading systems are machine learning algorithms, which are massive code libraries built at the intersection of artificial intelligence and statistics. Luckily for asset managers, the heavy lifting of applied mathematics and linear algebra has already been completed although academia is pushing these technologies further every day. Machine learning is already driving the spectacular advances in language translations, predictive consumer preferences, medical science and many other fields. Much of the software is open source so the real challenge for traders is to build and manage the databases that allow the software to make accurate predictions.

ALTERNATIVE DATA AND NATURAL LANGUAGE PROCESSING (NLP)

While computers operate at blindingly fast speeds, even a child is better at understanding spoken language and precocious first graders are far better readers. Computers require NLP to preprocess communication into digital form. During the past decade, new machine libraries like Google’s Tensor-Flow have transformed the world of NLP, powering remarkable technologies like Google Translate. NLP is now being used to listen-in on companies’ earning call, to translate regulatory findings, and process sentiment from social media. Predictive Technology Systems is using NLP to update its massive ML training sets every second of the trading day to identify and capitalize on new opportunities.

Predictive Technology Systems works with all of the major alternative data providers to provide its clients automated access to real time alternative data streams. Many of these vendors provide digital signals for interpreting the “sentiment” of their feeds. Or if you desire a proprietary NLP solution, we can help build one for you.

AUTOMATED TRADING

Once the trading systems is built and making predictions, to execute the trades one needs an automated connection to a licensed broker/dealer with a robust API (applications programming interface). At PTS, we currently use Interactive Brokers to execute trades. We are willing to follow our clients’ preference in connecting with other broker/dealers.

FEATURE ENGINEERING

Preparing a “training set” for processing by ML software requires constructing a large table of rows and columns similar in appearance to a very large Excel or spreadsheet file. Each row of data corresponds to each “tick” in an equity bid/ask price, and since our data tables use years of historical data, the number of rows is in the hundreds of millions. By convention, the first column in the spreadsheet is the “feature” that we’re trying to predict. In our case this is often a two minute or five minute change in the price of an equity. The remaining columns, often numbering in the hundreds, are the “features” that we suspect might have a correlation with the prediction column. Identifying and building these features is a non trivial task called “feature engineering.”