Skanalytix Pty Ltd

Overview

Despite the enormous recent progress in AI and machine learning (ML), the reality is that most organizations must deal with tabular data consisting of a mix of numeric and categorical attributes, often with highly skewed distributions and missing values. Such datasets continue to pose challenges, and it is in this space that we are focused.

Skanalytix has developed a new graph-based computational framework for ML named Unified Numerical-Categorical Representation and Inference (UNCRi), at the heart of which lies a unique data representation scheme coupled with a powerful inference procedure that can be used to estimate the probability distribution of any target variable, conditional on the values of one or more other variables. This flexibility allows a wide variety of generic data-oriented tasks to be performed. These include common tasks such as prediction (classification and regression), clustering and data imputation, but also far more challenging tasks such as full joint probability estimation and synthetic data generation. But these out-of-the-box solutions only scratch the surface. The framework’s breadth and flexibility allow it to easily extend to custom tasks such as the development of recommender systems.

We have just released a free online version of our new Synthetic Financial Time Series Generator, which generates realistic synthetic series that mimic the properties of real series. Not only do the generated series accurately capture stylized features such as fat tails, volatility clustering and mean reversion, but can also capture positive or negative correlations between multiple series. You can try it out here: https://skanalytix.com/synthetic-financial-time-series-generator.