Structured latent models improve multi-target time series forecasting
BeliefGraph-JEPA: Structured Latent World Models for Action-Conditioned Time Series
Machine LearningArtificial Intelligence
Summary
Predicting how different actions affect multiple things over time is tricky because influences can be delayed and last for different amounts of time. The authors developed BeliefGraph-JEPA, a system that models these influences as special hidden states that change over time and connect to individual targets through a graph. This method helps the system better understand and predict outcomes when multiple factors are involved. They tested their approach on four real-world systems like healthcare and agriculture, showing it forecasts better than existing methods.
What this means in practice
- •For hospital data teams: Improve predictions of patient health outcomes by modeling delayed and target-specific effects of treatments and interventions over time.
- •For agricultural monitoring teams: Enhance crop yield forecasts by capturing how weather and farming actions influence multiple environmental variables with varying delays.
Authors
Yue Li, Kangqi Ni, Zhen Tan, Tianlong Chen
Abstract
Action-conditioned time-series forecasting requires accounting for how future actions and exogenous forcings influence multiple targets through partially observed effects with different delays and persistence. Direct conditioning leaves the evolution and target-specific influence of these effects implicit in the predictor, while static relational graphs specify connections without tracking evolving effects. This motivates representing future-driver influence through structured latent states that evolve over the forecast horizon and route information to individual targets. We introduce BeliefGraph-JEPA, a structured latent world model that factorizes driver influence into typed latent-effect states. These states are rolled forward under future drivers and routed through a graph to target-specific nodes, forming the predictive base of a joint-embedding predictive architecture. A capacity-controlled residual supplements this base with direct driver information. On four multi-target clinical, agricultural, environmental, and industrial systems, the framework outperforms a range of pretrained and supervised known-future-covariate baselines. Matched controls isolate latent dynamics, future rollout, graph routing, and residual capacity; future rollout and graph-first residual routing improve forecasting across all four systems.