Incremental Learning of Stock Trends via Meta-Learning with Dynamic Adaptation

Forecasting the trend of stock prices is an enduring topic at the intersection of finance and computer science. Incremental updates to forecasters have proven effective in alleviating the impacts of concept drift arising from non-stationary stock markets. However, there is a need for refinement in the incremental learning of stock trends, as existing methods disregard recurring patterns. To address this issue, we propose meta-learning with dynamic adaptation (MetaDA) for the incremental learning of stock trends, which performs dynamic model adaptation considering both the recurring and emerging patterns. We initially organize the stock trend forecasting into meta-learning tasks and train a forecasting model following meta-learning protocols. During model adaptation, MetaDA adapts the forecasting model with the latest data and a selected portion of historical data, which is dynamically identified by a task inference module. The task inference module first extracts task-level embeddings from the historical tasks, and then identifies the informative data with a task inference network. MetaDA has been evaluated on real-world stock datasets, achieving state-of-the-art performance with satisfactory efficiency.