Japanese Trading Analytics

Retail investor clickstream data can add value in a variety of investment use cases, ranging from pre-trade analytics to short-term investment strategies. However, processing any voluminous clickstream feed necessary to leverage such data in production presents many challenges.

<p><br/></p>

Our time-series attention signal (TSAS) solves the challenges of voluminous clickstream data by applying novel statistical techniques to the clickstream data, resulting in analytics that can easily be integrated into investment processes. We apply a MSTL model (Multiple Seasonal-Trend decomposition using LOESS) is to clickstream data from Japanese retail stock research websites to identify anomalies in the volume of clicks at the specific company level. The model is updated weekly, ensuring that recent trends in browsing activity are incorporated into the model.

Model outputs are delivered hourly throughout the trading day, from 8:30am JST until 2:30pm JST, together with descriptive statistics about underlying clickstream data volumes. The result is a steady heartbeat of predictive, actionable information to market participants throughout the trading day.

<p><br/></p>