Inequality Metrics as Early Warning Indicators of Bitcoin Market Instability

Abstract

Existing Bitcoin price prediction systems mainly rely on traditional statistical models or basic financial indicators, which often fail to capture complex market patterns and the highly volatile nature of cryptocurrency markets. Many machine learning models use only historical price values and do not integrate statistical inequality measures such as the Lorenz Curve, Gini Index, or K-Value, which can provide insights into market distribution and stability. To overcome these limitations, the proposed system analysis historical Bitcoin price data from 2010 to 2021, including the open, high, low, and closing prices. The proposed framework incorporates the Lorenz Curve, Gini Index, and K-Value along with financial data, and a linear regression model is applied for time-series forecasting. The proposed model showed improvements of 1.46%, 6.87%, and 0.07% in RMSE, MAE, and MAPE, respectively, over the baseline. The results indicate that integrating inequality-based features improved prediction reliability and provided a better understanding of market behaviour.

Introduction

Bitcoin is the first and one of the most widely used cryptocurrencies, operating through a decentralized network based on blockchain technology. Since its introduction, Bitcoin has gained popularity as a digital asset and investment option. However, its highly volatile prices, influenced by market demand, investor sentiment, and global economic conditions, make accurate price prediction challenging. Machine learning and statistical analysis can help identify historical patterns and forecast potential price movements.

The main problem addressed in this research is the limitation of existing Bitcoin price prediction models, which primarily rely on historical prices and conventional financial indicators. These models may struggle to capture nonlinear market behaviour, extreme price fluctuations, and the distribution of market changes. Therefore, the study proposes a hybrid prediction framework that combines machine learning algorithms with statistical inequality measures, including the Lorenz Curve, Gini Index, and Kolkata (K) Index, to analyse price fluctuations and improve forecasting performance.

The literature review examines various approaches, including Long Short-Term Memory (LSTM) networks, ensemble learning, Transformers, reinforcement learning, sentiment analysis, and regression models. Although these techniques offer advantages in forecasting, they may involve high computational costs, require large datasets, or overlook statistical inequality measures. This research aims to address the gap by integrating inequality analysis with machine learning-based Bitcoin price prediction.

The proposed methodology uses historical Bitcoin market data from 2010 to 2021, containing daily opening, closing, highest, and lowest prices, trading volume, dates, and percentage changes. The dataset is preprocessed by handling missing values, removing duplicates, and applying normalization and feature scaling. Additional statistical features, including the Lorenz Curve, Gini Index, Kolkata Index, mean, standard deviation, volatility, and price changes, are then derived. The dataset is divided into 80% training data and 20% testing data.

The proposed system evaluates machine learning algorithms such as Linear Regression, Random Forest Regression, Decision Tree Regression, and K-Nearest Neighbour (KNN). Model performance is assessed using Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and the coefficient of determination (R2R2). The best-performing model is selected based on its prediction accuracy and error metrics. The system also aims to support market trend analysis, identify potential instability, and provide useful information for investors and financial analysts.

The reported results compare the proposed model with a baseline model. The proposed approach demonstrates improvements of 1.46% in RMSE, 6.87% in MAE, and 0.07% in MAPE. These findings suggest that incorporating statistical inequality measures may improve certain aspects of Bitcoin price prediction, although further validation is needed to establish the generalizability of the results.

Conclusion

In this project, a hybrid framework for Bitcoin price prediction was developed by combining machine learning methods with statistical inequality analysis. The main goal of this approach was to improve the accuracy and reliability of cryptocurrency price forecasting while overcoming the limitations of traditional prediction techniques. The proposed system used historical Bitcoin price data along with important financial indicators such as opening price, closing price, highest price, and lowest price. In addition to these standard features, statistical inequality measures like the Lorenz Curve, Gini Index, and K-Value were included to study the distribution and stability of price variations. These measures helped provide deeper insights into market volatility patterns that are usually not captured by conventional prediction models. A Linear Regression algorithm was applied as the machine learning model to perform time-series forecasting. The model was trained and tested using preprocess historical data, and its performance was evaluated using suitable accuracy metrics. The experimental results showed that incorporating inequality-based features improved prediction reliability and helped in better understanding market behaviour. The study also emphasized the importance of integrating statistical analysis with machine learning techniques in order to capture both distribution characteristics and predictive trends. This hybrid approach is more effective in handling the dynamic and highly volatile nature of cryptocurrency markets compared to traditional models.

References

[1] Ahmed M, Khan S and Hussain T (2024), “Bitcoin price prediction using deep reinforcement learning”, Information Sciences, Vol. 650, pp. 203-218. [2] Chen L and Zhao Y (2024), “Transformer-based time series prediction for cryptocurrency markets”, Neural Computing and Applications, Vol. 36, pp. 4571-4585. [3] Gupta A, Verma S and Joshi R (2026), “Time series prediction of Bitcoin using advanced regression models”, Computational Economics, Vol. 67, pp. 789-804. [4] Kim J and Park S (2023), “Hybrid machine learning model for cryptocurrency forecasting”, Expert Systems with Applications, Vol. 214, pp. 119-132. [5] Kumar N and Gupta R (2024), “Statistical inequality analysis of cryptocurrency markets”, Journal of Economic Model, Vol. 128, pp. 112-125. [6] Li Q, Sun X and Wang J (2024), “Multivariate cryptocurrency prediction using machine learning”, IEEE Transactions on Big Data, Vol. 10, No. 2, pp. 789-801. [7] Manna S S, Biswas S and Chakrabarti B K (2022), “Near universal values of social inequality indices in self-organized critical models”, Physica A: Statistical Mechanics and its Applications, Vol. 596, p. 12721. [8] Patel D, Shah P and Mehta K (2025), “Bitcoin forecasting using sentiment analysis and machine learning”, Applied Intelligence, Vol. 55, pp. 311-326. [9] Rahman M and Islam S (2026), “Inequality-driven machine learning framework for financial market prediction”, Journal of Financial Data Science, Vol. 8, No. 1, pp. 44-58. [10] Santos F, Pereira M and Silva R (2025), “Hybrid deep learning framework for cryptocurrency forecasting”, Future Generation Computer Systems, Vol. 148, pp. 99-112. [11] Singh R, Kumar A and Sharma V (2023), “Bitcoin price prediction using ensemble learning techniques”, Applied Soft Computing, Vol. 132, pp. 109-121. [12] Wang T and Zhou Y (2025), “Explainable AI for cryptocurrency price prediction”, Knowledge-Based Systems, Vol. 294, pp. 111-124. [13] Zhang Y, Liu H and Wang P (2023), “Deep learning-based cryptocurrency price prediction using LSTM networks”, IEEE Access, Vol. 11, pp. 34567-34579. [14] Zhou X, Chen H and Li D (2025), “Graph neural networks for cryptocurrency market prediction”, IEEE Transactions on Neural Networks and Learning Systems, Vol. 36, No. 5, pp. 1567-1579.

Copyright

Copyright © 2026 U V V Rama Mohana Krishna Sai, Dr. K V Ramana. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.