Modeling Fuzziness and Randomness with Cloud Models

20 concepts12 papers1 proposal

When evaluating high-stakes engineering or environmental systems—such as the safety of a deep foundation pit (Wu et al.) or the risk of a dam failure (Zhang)—decision-makers face two fundamentally different types of uncertainty: fuzziness and randomness.

  • Fuzziness refers to boundary vagueness. When an expert describes a slope as "moderately stable," the boundary between stable and unstable is fuzzy.
  • Randomness refers to probabilistic variability. Sensor readings monitoring a tunnel fluctuate continuously due to environmental noise and measurement errors.

Traditional frameworks struggle to handle both simultaneously. Fuzzy set theory handles vagueness well but assigns deterministic membership degrees (e.g., a fixed membership of 0.7), ignoring stochastic volatility. Probability theory excels at randomness but requires crisp, well-defined event boundaries.

The Cloud Model bridges this gap. As Han et al. describe, it is an algorithmic approach that synthesizes fuzzy mathematics and probability theory to map qualitative concepts into quantitative data. Instead of a rigid membership curve, it represents information as a cluster of stochastic points called cloud droplets (Tang et al.), capturing both the vagueness of human thought and the randomness of physical data.


The Three Digital Characteristics

To mathematically define a qualitative concept without stripping away its inherent uncertainty, the cloud model condenses the concept into three core parameters: Expectation ($Ex$), Entropy ($En$), and Hyper-entropy ($He$).

       ^  Membership Degree (μ)
       |          ..''``..         <- Cloud Droplets
    1  |        .' . :. . '.
       |       :  .  :  .  :
       |      ' .  . : .  . '
    0  +------'------:------'------> Quantitative Value
                    Ex
              |<--- En --->|       (Width of the cloud)
               Dispersion = He   (Thickness/blurriness of the curve)

1. Expectation ($Ex$)

Expectation is the central point of the cloud. It represents the most typical, mathematically representative value of the qualitative concept. For instance, if experts agree that a temperature of 25°C is the absolute ideal definition of "comfortable," then $Ex = 25$.

2. Entropy ($En$)

Entropy measures the breadth or coverage of the concept, quantifying its fuzziness. A larger $En$ means the concept is more vague and spans a wider range of acceptable values. As Fan Yang notes, entropy dictates how far a value can stray from the expectation while still belonging to the qualitative concept.

3. Hyper-entropy ($He$)

Hyper-entropy is the uncertainty of the entropy itself, quantifying randomness. It determines how dispersed and scattered the cloud droplets are. If $He = 0$, the cloud collapses into a traditional, smooth fuzzy membership curve. When $He > 0$, the membership degree of any given point fluctuates stochastically. Qiuwen Zhang explains that $He$ accounts for the fact that different experts might disagree on the boundaries of a concept, or that data noise makes the membership degree itself random.


Why It Works: Preserving the Gray Areas

Traditional risk assessment methods often normalize continuous data into rigid, deterministic intervals (e.g., mapping a safety index strictly between 0 and 1). However, Wang et al. point out that this rigid normalization masks critical risks by ignoring the "gray areas" where safety levels overlap.

The cloud model works because it avoids binary or fixed-boundary classifications. By generating thousands of cloud droplets, it visualizes risk as a probability distribution rather than a single deterministic score (Xiong et al.; Gong et al.). If a sensor reading falls near a dangerous threshold, a traditional model might flag it as safe if it sits just below the line. A cloud model, by contrast, will generate droplets that span into the "danger" zone, revealing a probabilistic risk of failure even if the expected value appears secure.


The Mathematical Engine: Cloud Generators

To transition between abstract human language and raw physical data, the framework relies on two directional algorithms known as cloud generators (Huang et al.).

Forward Cloud Generator (Concept to Data)

The forward cloud generator takes the three digital characteristics ($Ex, En, He$) and produces a specific number of cloud droplets (Wu et al.).

  1. It generates a normal random variable $En'$ with a mean of $En$ and a variance of $He^2$.
  2. It generates a specific data point $x$ with a mean of $Ex$ and a variance of ${En'}^2$.
  3. It calculates the membership degree $\mu$ based on a Gaussian distribution.

By repeating this process, a visual "cloud" emerges, transforming a qualitative concept into a probabilistic membership function.

Inverse Cloud Generator (Data to Concept)

The inverse cloud generator works in reverse. It takes a collection of raw, noisy data points (such as historical pipeline failure metrics or grid operation logs) and extracts the underlying $Ex, En,$ and $He$ (Gong et al.). This allows researchers to mathematically formalize real-world data into structured, uncertain concepts that can be easily compared against safety standards.


Real-World Integration

In practice, cloud models are rarely used in isolation; they serve as the uncertainty-handling core of broader decision-making frameworks.

They are frequently paired with weighting schemes like the Analytic Hierarchy Process (AHP) or the Entropy Weight Method to evaluate complex, multi-variable systems. For example, the authors have successfully integrated cloud models to assess the comprehensive benefits of power transmission networks (Tang et al.), the investment efficiency of pumped storage power plants (Lü et al.), and high-dimensional slope stability data (Wang et al.). By replacing crisp numbers with cloud parameters, these hybrid frameworks ensure that subjective expert opinions and objective data are fused without losing the critical nuances of uncertainty.

History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_cloud_model_89fb66dd,
  title = {cloud model},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/cloud-model/rdr_89fb66dd7ead43b890cdf2b285a6165d},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-22},
}

This exact version (version 1, frozen text):

@misc{lacuna_cloud_model_89fb66dd_v1,
  title = {cloud model},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/cloud-model/rdr_89fb66dd7ead43b890cdf2b285a6165d/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-22},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake