Time Series Forecasting
OVERVIEW
Our Predictive Analytics Engine TSF(Time Series Forecasting), based on advanced deep learning technology, analyses large amounts of historical data of different kinds to make accurate forecasts. The engine is particularly great at detecting long and short term patterns where standard statistical approaches usually fail.
FEATURES
-
Realtime
- The TSF engine is capable of ingesting large amounts of high dimensional data in realtime and produces forecasts relatively quickly, making it suitable for applications where speed is a constraint.
-
Continuous learning
- The algorithms powering TSF are constantly learning from the new data they receive, hence, are able to output up-to-date and accurate forecasts even in the case of changing patterns in the data.
-
Customizable
- TSF is easily customisable for different forecasting time-horizons (daily, intraday...) and other specific requirements, such as training on proprietary data.
CASES
-
Trade Volume Prediction（大和証券）
Daiwa Securities Co. Ltd.
-
Background
Large financial institutions trade stock on behalf of their clients, and they want to execute in the best possible way.
-
Problem
Financial time series are notoriously difficult to forecast and traditional statistical models often make large errors.
-
Result
Using TSF - a deep learning based forecasting engine - the volume predictions are more accurate.
-