Towards Dynamic Trend Filtering through Trend Point Detection with Reinforcement Learning

2024-06-06

International Joint Conference on Artificial Intelligence

AI Summary

Traditional trend filtering relies on global smoothness assumptions, making it difficult to identify and preserve abrupt change points in time series—often conflating critical turning points with noise. To address this, we propose the first reinforcement learning–based trend point detection framework, reformulating trend filtering as a Markov Decision Process (MDP). We introduce the concept of Dynamic Trend Points (DTPs), enabling local adaptivity of smoothness according to heterogeneous data characteristics. Our method employs a discrete-action-space policy, a forecast-oriented sum-of-squares loss reward function, and DTP-driven dynamic interpolation for trend reconstruction. Evaluated on multiple real-world and synthetic datasets, our approach achieves significantly improved change-point detection accuracy and preserves fidelity of key subsequences in trend estimation. Downstream forecasting errors decrease by an average of 12.7% compared to state-of-the-art methods.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Reasoning under Uncertainty: Sequential Decision Making
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Trend filtering simplifies complex time series data by applying smoothness to filter out noise while emphasizing proximity to the original data. However, existing trend filtering methods fail to reflect abrupt changes in the trend due to `approximateness,' resulting in constant smoothness. This approximateness uniformly filters out the tail distribution of time series data, characterized by extreme values, including both abrupt changes and noise. In this paper, we propose Trend Point Detection formulated as a Markov Decision Process (MDP), a novel approach to identifying essential points that should be reflected in the trend, departing from approximations. We term these essential points as Dynamic Trend Points (DTPs) and extract trends by interpolating them. To identify DTPs, we utilize Reinforcement Learning (RL) within a discrete action space and a forecasting sum-of-squares loss function as a reward, referred to as the Dynamic Trend Filtering network (DTF-net). DTF-net integrates flexible noise filtering, preserving critical original subsequences while removing noise as required for other subsequences. We demonstrate that DTF-net excels at capturing abrupt changes compared to other trend filtering algorithms and enhances forecasting performance, as abrupt changes are predicted rather than smoothed out.

Problem

• Detecting abrupt trend changes in time series data
• Overcoming constant smoothness in existing filtering methods
• Preserving critical subsequences while filtering noise dynamically

Innovation

• Reinforcement Learning for Dynamic Trend Points
• Markov Decision Process models trend detection
• DTF-net flexibly filters noise, preserves changes