Unlocking Insights with Grey Relational Analysis and Grey Prediction Models

When dealing with complex systems or problems, data is often limited, especially in engineering, economics, and environmental fields.

Grey theory , a method for handling uncertainty, fuzziness, and incomplete data, has been widely applied in these areas.

Grey Relational Analysis (GRA) and Grey Prediction Model (GPM) are two important components of grey theory. By effectively analyzing and forecasting limited data, they provide strong support for decision makers. This article explores the basic principles and applications of GRA and GPM.

1. Grey Relational Analysis

1.1 Basic Concept of Grey Relational Analysis

Grey Relational Analysis is a mathematical method that reveals relationships between multiple systems (or subsystems) by comparing their relational coefficients. Unlike traditional statistical methods, it does not require complete statistical properties or full information; it extracts valuable information from existing data and adapts well to incomplete information.

The core idea is to measure similarity between two sequences using a "relational degree"; a larger degree indicates a closer relationship.

1.2 Calculation of Grey Relational Degree

The calculation of the grey relational degree is the core step of GRA. Suppose there are two sequences representing historical data of two systems.

1. Generation of grey relational sequences : Standardize the original data to eliminate dimensional effects. The standardization formula is applied.

2. Compute absolute differences : Calculate the absolute differences between the two standardized sequences.

3. Calculate grey relational degree : Use the following formula (where the distinguishing coefficient is typically between 0.5 and 1.0) to compute the relational degree.

A larger grey relational degree indicates a tighter relationship between the sequences.

1.3 Applications of Grey Relational Analysis

Grey Relational Analysis is widely used in many fields, especially social economics, environment, and engineering management.

Indicator selection: Identify key indicators most related to the target from multiple metrics.

System evaluation: Assess the impact of various factors on system performance.

Trend prediction: Reveal potential trends in historical data to guide future directions.

2. Grey Prediction Model

2.1 Basic Concept of Grey Prediction Model

The Grey Prediction Model is a dynamic forecasting method based on grey system theory, suitable for situations with missing data or incomplete information. It builds a model from historical data to predict future system trends. The most common model is the GM(1,1) model.

GM(1,1) is the basic single-variable first-order grey model; the "1,1" denotes a single variable.

2.2 Modeling Process of GM(1,1) Model

Accumulated generation of original data : Perform a cumulative sum on the original data series to obtain the accumulated series.

Establish grey differential equation : Based on the accumulated series, set up a first-order grey differential equation.

Solve parameters : Estimate the model parameters using the least squares method so that the accumulated series fits the model closely.

Build prediction model : Construct the prediction model using the estimated parameters.

Restore predicted data : Convert the predicted values of the accumulated series back to the original data scale to obtain forecasts.

2.3 Applications of GM(1,1) Model

The grey prediction model is widely applied to real-world problems, especially where data is scarce and uncertainty is high.

Economic forecasting: Predict national economy, industry development, and enterprise performance.

Environmental forecasting: Long‑term prediction of pollution, climate change, and other environmental trends.

Engineering management: Forecast production processes, project schedules, and quality control.

3. Combining Grey Relational Analysis and Grey Prediction Model

Grey Relational Analysis and Grey Prediction Model can be combined to enhance analysis precision and prediction accuracy. In practice, GRA can identify key factors influencing system development, while the prediction model forecasts future trends based on those factors.

For example, in enterprise sales forecasting, GRA helps identify influential factors such as price, advertising spend, and competitors, and the grey prediction model uses these factors to predict future sales.