Vista Analytics Announces Improved Machine Learning Methodology to Address Issues of Concept Drift in Legal Disputes

Share Article

Innovative Legal Technology Company Continues to Drive Improvements for Leveraging Artificial Intelligence for Discovery

Vista Analytics has announced improvements to its machine learning solutions, which address the issues of concept drift during discovery. This past December, Vista data scientist Yihua Shi Astle presented a paper at the 2017 IEEE International Conference on Big Data. Vista's submission-co-authored by Yihua, Dr. Xuning Tang, and co-founder Craig Freeman-was entitled "Application of Dynamic Logistic Regression with Unscented Kalman Filter in Predictive Coding" and was selected for presentation by some of the most respected data scientists and PhDs. in the world and is a testament to Vista's groundbreaking work in litigation technology research. Since the publication of the paper, Vista's team of data scientists has developed key improvements to the methodology that not only provide more accurate results, but do so more efficiently thus reducing a project's timeline.

Astle says, "Since the publication of the paper, we have further improved our predictive coding methodology by harnessing the power of ensemble learning. It increases the stability of our previous approach, while retaining its ability to adapt to new data quickly. With ensemble learning, we now can take advantage of many different classification algorithms, and we have seen great improvement in our model accuracy."

Freeman adds, "The issue of concept drift during discovery is not being widely addressed in our industry. The only thing that can be guaranteed in a large matter is that there are going to be changes to the underlying data set over time and that the attorneys will need to look for new issues as they become more familiar with the matter. If you are not addressing these difficult problems in predictive coding from the outset, you are working from a great disadvantage. By factoring in concept drift our approach learns new patterns at a faster rate, renders better accuracy and recall, and requires a reduced labeling cost, which when combined makes it potentially a superior alternative in updating predictive coding models."

Vista Analytics continues to strongly believe that the best way to incorporate machine learning within litigation is to take advantage of the rapid advances in the field and not attempt to productize one model that can become obsolete before it makes it to market. Vista will continue to constantly adjust and improve its platform using the cutting edge machine learning tools, which are rapidly improving.

To read the paper and see the presentation given at the IEEE Big Data Conference, please visit

About Vista Analytics

Vista Analytics is harnessing machine learning, artificial intelligence, and the scalability of a cloud-based infrastructure to enhance client satisfaction by reducing costs and timelines, all while providing valuable insights beyond what current technologies provide. Founded by Craig Freeman and Michael Faraci, Vista Analytics has been expanding its team to include world-class talent comprised of PhD level researchers and analysts. The company is located at 1333 H Street, N.W., Suite 600W, Washington, D.C. 20005. Vista Analytics can be reached at (202) 808-2286 or through its website,


Share article on social media or email:

View article via:

Pdf Print

Contact Author

Michael Faraci
Vista Analytics
+1 (202) 262-4885
Email >

Craig Freeman
Visit website