Abstract:The fuzzy cognitive map (FCM) is a powerful model for system state prediction and interpretable knowledge representation. Recent years have witnessed the tremendous efforts devoted to enhancing the basic FCM, such as introducing temporal factors, uncertainty or fuzzy rules to improve interpretation, and introducing fuzzy neural networks or wavelets to improve time series prediction. But how to achieve high-precision yet interpretable prediction in cross-domain real-life applications remains a great challenge. In this article, we propose a novel FCM extension called deep FCM (DFCM) for multivariate time series forecasting, in order to take both the advantage of FCM in interpretation and the advantage of deep neural networks in prediction. Specifically, to improve the predictive power, DFCM leverages a fully connected neural network to model connections (relationships) among concepts in a system, and a recurrent neural network to model unknown exogenous factors that have influences on system dynamics. Moreover, to foster model interpretability encumbered by the embedded deep structures, a partial derivative-based approach is proposed to measure the connection strengths between concepts in DFCM. An alternate function gradient descent algorithm is then proposed for parameter inference. The effectiveness of DFCM is validated over four publicly available datasets with the presence of seven baselines. DFCM indeed provides an important clue to building interpretable predictors for real-life applications.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to achieve high - precision and interpretable multivariate time - series prediction in cross - domain practical applications. Specifically, the paper points out that although the traditional Fuzzy Cognitive Map (FCM) performs well in system - state prediction and interpretable knowledge representation, its ability to handle nonlinear - dynamics prediction is limited, especially when facing unknown exogenous factors in complex real - world applications. Moreover, although deep - learning models perform excellently in prediction performance, they are usually regarded as black - box models and are difficult to interpret. Therefore, how to improve the nonlinear - dynamics prediction ability of FCM while maintaining its interpretive advantages, or in other words, achieve satisfactory interpretable prediction in practical applications, remains an open and important problem. To meet this challenge, the paper proposes a new FCM - extended model - Deep Fuzzy Cognitive Map (Deep FCM), aiming to combine the interpretive ability of FCM and the prediction ability of deep neural networks to achieve high - precision and interpretable multivariate time - series prediction. Specific improvements include: 1. **Improving prediction ability**: Deep FCM uses a fully - connected neural network to model the relationships between system concepts and uses a Recurrent Neural Network (RNN) to model unknown exogenous factors that have an impact on system dynamics. 2. **Enhancing model interpretability**: To overcome the interpretability problems brought by the deep structure, the paper proposes a method based on partial derivatives to measure the connection strength between concepts in Deep FCM. 3. **Parameter inference algorithm**: An Alternate Function Gradient Descent algorithm is designed for parameter inference to ensure the effectiveness and accuracy of the model. Through these improvements, Deep FCM can verify its effectiveness on multiple public data sets and compare with seven baseline methods. The results show that Deep FCM can indeed achieve better performance in system - state prediction while maintaining the interpretability of the FCM framework.

Deep Fuzzy Cognitive Maps for Interpretable Multivariate Time Series Prediction

Application Research in Stock Market Forecast with Fuzzy Cognitive Maps

Time Series Prediction Using Sparse Autoencoder and High-Order Fuzzy Cognitive Maps

Convex Optimization Based High-Order Fuzzy Cognitive Map Modeling and Its Application in Time Series Predicting

Learning High-Order Fuzzy Cognitive Maps Via Multimodal Artificial Bee Colony Algorithm and Nearest-Better Clustering: Applications on Multivariate Time Series Prediction

Is Single Enough? A Joint Spatiotemporal Feature Learning Framework for Multivariate Time Series Prediction

Double Decomposition and Fuzzy Cognitive Graph-Based Prediction of Non-Stationary Time Series

Time-series forecasting based on fuzzy cognitive maps and GRU-autoencoder

Time series forecasting based on kernel mapping and high-order fuzzy cognitive maps

Multiple-Input Multiple-Output Randomized Fuzzy Cognitive Map Method for High-Dimensional Time Series Forecasting

Time Series Prediction Based on LSTM and High-Order Fuzzy Cognitive Map with Attention Mechanism

Fuzzy-clustering and fuzzy network based interpretable fuzzy model for prediction

Training Novel Adaptive Fuzzy Cognitive Map by Knowledge-Guidance Learning Mechanism for Large-Scale Time-Series Forecasting

CNN-FCM: System modeling promotes stability of deep learning in time series prediction

Differential Convolutional Fuzzy Time Series Forecasting

Compression and regularized optimization of modules stacked residual deep fuzzy system with application to time series prediction

Multivariate Time Series Forecasting Based on Elastic Net and High-Order Fuzzy Cognitive Maps: A Case Study on Human Action Prediction Through EEG Signals

Classification and Feature Transformation with Fuzzy Cognitive Maps

A Fuzzy C-means Clustering-Based Hybrid Multivariate Time Series Prediction Framework with Feature Selection

A Novel Framework for Enhanced Interpretability in Fuzzy Cognitive Maps

Self-Constructed Deep Fuzzy Neural Network for Traffic Flow Prediction