Fuzzy Time Series Forecasting Models and Applications

Fuzzy time series forecasting has emerged as a versatile methodology for predicting complex and uncertain temporal data by translating crisp observations into linguistic variables. Early approaches established the basic framework of fuzzification, rule induction and defuzzification, which enabled robust handling of imprecision in domains such as enrolment projections and financial indices. Contemporary research has refined interval partitioning through adaptive clustering, genetic algorithms and tree-based schemes to optimise the division of the universe of discourse. Extensions into intuitionistic and neutrosophic fuzzy sets introduce degrees of non-membership and indeterminacy, enriching the representation of ambiguity. Hybrid models that integrate fuzzy inference with neural networks, Markov chains and singular spectrum analysis have further enhanced forecast accuracy by capturing both linear and nonlinear dependencies. Applications now span air pollution monitoring, energy load prediction, stock market analysis and supply-chain planning, demonstrating global significance in resource management and risk assessment. Persistent challenges include determining optimal membership functions, selecting appropriate order and interval lengths, and ensuring generalisation across diverse datasets. Recent advances address these by automating parameter tuning and combining probabilistic transitions with fuzzy logic, thus offering more reliable and interpretable forecasting frameworks for decision support in uncertain environments.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Fuzzy Time Series Forecasting Models and Applications publication trend

The graph below shows the total number of articles in fuzzy time series forecasting models and applications across all publications each year (not limited to Nature Index journals).

Technical terms

Fuzzy time series: A forecasting methodology that maps numerical time series into fuzzy sets and derives linguistic relationships for prediction.

Fuzzification: The process of converting precise numerical data into fuzzy linguistic terms via membership functions.

Membership function: A mathematical curve assigning each datum a degree of membership between zero and one within a fuzzy set.

Intuitionistic fuzzy set: An extension of classical fuzzy sets characterised by degrees of membership, non-membership and hesitancy.

Markov chain: A stochastic model representing transitions between discrete states based on conditional probabilities.

Partitioning: Division of the data universe into intervals or fuzzy sets that underpin rule induction and inference.

References

1. A New Bandwidth Interval Based Forecasting Method for Enrollments Using Fuzzy Time Series. Applied Mathematics (2011).
2. A New Time‐Invariant Fuzzy Time Series Forecasting Method Based on Genetic Algorithm. Advances in Fuzzy Systems (2012).
3. A Novel Fuzzy Time Series Forecasting Model Based on Multiple Linear Regression and Time Series Clustering. Mathematical Problems in Engineering (2020).
4. A Novel Stochastic Fuzzy Time Series Forecasting Model Based on a New Partition Method. IEEE Access (2021).
5. Predicting Daily Air Pollution Index Based on Fuzzy Time Series Markov Chain Model. Symmetry (2020).
6. Intuitionistic fuzzy time series functions approach for time series forecasting. Granular Computing (2020).
7. A Refined Approach for Forecasting Based on Neutrosophic Time Series. Symmetry (2019).
8. Time series forecasting using singular spectrum analysis, fuzzy systems and neural networks. MethodsX (2020).