|
Abstract:
|
Recent advances in data collection and data storage technologies substantially increased volumes and varieties of available time-related data. To analyze these data, IBM products include technologies that handle three kinds of time-related data: regular time series, spatiotemporal, and transactional data. The technologies were developed in collaboration with IBM Research. This workshop will introduce participants to the theory of each algorithm and demonstrate their use in IBM products with real-life data. For regular time series data, Temporal Causal Modeling captures causal relationships between variables in datasets with large numbers of time series variables, and performs forecasting for future values. Goal seek analysis controls target values in the future by optimizing the values of predictors. Spatiotemporal prediction models forecast future target values based on models incorporating both geographic and temporal information. For transactional data, Event-Based Time Series helps users discover sequence rules between events incorporating time durations between events and the values of the events. These methods will be demonstrated in IBM SPSS Modeler and IBM SPSS Statistics products, as well as the new cloud-based IBM Data Science Experience. Basic knowledge of time series analysis and data mining is assumed.
|
