Paper: arXiv 2409.06728
Authors: Sanjay Sathish, Charu C Sharma
Abstract
Our research presents a new approach for forecasting the synchronization of stock prices using machine learning and non-linear time-series analysis. To capture the complex non-linear relationships between stock prices, we utilize recurrence plots (RP) and cross-recurrence quantification analysis (CRQA). By transforming Cross Recurrence Plot (CRP) data into a time-series format, we enable the use of Recurrent Neural Networks (RNN) and Long Short-Term Memory (LSTM) networks for predicting stock price synchronization through both regression and classification. We apply this methodology to a dataset of 20 highly capitalized stocks from the Indian market over a 21-year period. The findings reveal that our approach can predict stock price synchronization, with an accuracy of 0.98 and F1 score of 0.83 offering valuable insights for developing effective trading strategies and risk management tools.
Complexity vs Empirical Score
- Math Complexity: 7.0/10
- Empirical Rigor: 8.5/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper employs advanced mathematics like cross-recurrence quantification analysis (CRQA) and recurrence plots (RPs), involving matrix theory and nonlinear dynamics. It demonstrates high empirical rigor by applying the methodology to a large, specific dataset (20 stocks over 21 years) with clear train-test splits, reporting concrete performance metrics (accuracy 0.98, F1 0.83) suitable for backtesting.
Research Flowchart
flowchart TD
A["Research Goal<br>Predict Stock Synchronization"] --> B["Data Input<br>20 Stocks x 21 Years"]
B --> C["Methodology<br>CRQA & Recurrence Plots"]
C --> D["Data Transformation<br>CRP to Time-Series"]
D --> E["Computational Process<br>RNN/LSTM Models"]
E --> F{"Classification"}
F --> G["Outcomes<br>Accuracy: 0.98, F1: 0.83<br>Trading Strategies & Risk Management"]
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes a methodology that uses distances from cross-recurrence plots as inputs to RNN and LSTM models to predict stock price synchronization.
Source sentence
Motivated by the earlier works in this paper we present a methodology for predicting the synchronization of stock movements. Our method considers the actual distances between the time points in the phase diagram instead of CRPs and takes these as an input to Recurrent Neural Network (RNN) and Long Short-Term Memory (LS
Assumes 4
- Adjusted daily stock price and volume series are normalized using training set mean and standard deviation. data
Source sentence
Each of the 𝑘(𝑘= 1,2, … 20) training time series was normalized by subtracting their respective mean𝑠 𝜇𝑘 and dividing by their respective standard deviations 𝜎𝑘.
- The first 70% of each pair’s data is used for training and the remaining 30% for testing. execution
Source sentence
we consider first 70% of the data (~3648 data points) as training price data to build the model and next 30% of the data (~1563 data points) as the testing dataset
- Missing data points lead to removal of six stocks, leaving 20 stocks for analysis. data
Source sentence
due to a significant number of consecutive missing data points, we had to remove 6 stocks from our analysis. Table 1 lists the details of the remaining 20 stocks.
- The CRP distance matrix can be reshaped into sub‑matrices (blocks) for regression forecasting. model
Source sentence
For each pair of stocks, we converted the distance matrix corresponding to the CRP into a format suitable for regression analysis, we take sub‑matrices along the diagonal of size 𝑛×𝑛 from the distance matrix and call them as a block.
Data
- Adjusted daily stock price and volume data market-prices public
Source sentence
... The time period considered for the analysis is from January 2003 to December 2023, a total of 5211 days.
Costs
Not applicable
Finds 3
- The proposed RNN/LSTM approach achieves an accuracy of 0.98 and an F1 score of 0.83 in predicting stock price synchronization. predictive-accuracy
Source sentence
The findings reveal that our approach can predict stock price synchronization, with an accuracy of 0.98 and F1 score of 0.83
- A CNN model using sub‑sampled CRP matrices underperformed with 55.47% accuracy and 0.45 F1 score. performance
Source sentence
this method underperformed on our dataset with accuracy of 55.47% and F1 score of 0.45.
- LSTM models generally outperform RNNs, especially with larger block sizes, showing higher R² and lower error metrics. performance
Source sentence
It can be inferred from table 1 that the LSTM model generally perform better than RNNs, particularly with larger block sizes of 20-30, as evidenced by slightly higher R² values and lower error metrics.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 21 related papers). Coverage is partial, so this is neither replication nor refutation.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.
Cited References (30)
- Recurrence plots for the analysis of complex systems (2025) — link
- Predicting the state of synchronization of financial time series using cross recurrence plots (2022) — link
- Neural Networks for Financial Time Series Forecasting (2022) — link
- Predicting LQ45 financial sector indices using RNN-LSTM (2021) — link
- Study on Accuracy Metrics for Evaluating the Predictions of Damage Locations in Deep Piles Using Artificial Neural Networks with Acoustic Emission Data (2021) — link
- Recurrence Plot and Machine Learning for Signal Quality Assessment of Photoplethysmogram in Mobile Environment (2021) — link
- Deep Learning-Based Weather Prediction: A Survey (2021) — link
- Financial Distress Prediction Using Hybrid Machine Learning Techniques (2020) — link
and 22 more
- Time-driven feature-aware jointly deep reinforcement learning for financial signal representation and algorithmic trading (2020) — link
- Deep Learning with Long Short-Term Memory Networks for Air Temperature Predictions (2019) — link
- Mutual information based stock networks and portfolio selection for intraday traders using high frequency data: An Indian market case study (2019) — link
- Recurrent Neural Networks for Financial Time-Series Modelling (2018) — link
- Classification of Recurrence Plots’ Distance Matrices with a Convolutional Neural Network for Activity Recognition (2018) — link
- Development of stock correlation networks using mutual information and financial big data (2018) — link
- Nonlinear dependencies on Brazilian equity network from mutual information minimum spanning trees (2017) — link
- Multidimensional Recurrence Quantification Analysis (MdRQA) for the Analysis of Multidimensional Time-Series: A Software Implementation in MATLAB and Its Application to Group-Level Data in Joint Action (2016) — link
- Anomaly detection in ECG time signals via deep long short-term memory networks (2015) — link
- Using the R-MAPE index as a resistant measure of forecast accuracy. (2013) — link
- Recurrence quantification analysis of global stock markets (2011) — link
- Collective behavior of stock price movements in an emerging market. (2007) — link
- Cross recurrence plot based synchronization of time series (2002) — link
- The false nearest neighbors algorithm: An overview (1997) — link
- Recurrence Plots of Dynamical Systems (1987) — link
- Forecasting The Air Temperature at a Weather Station Using Deep Neural Networks (2020) — link
- Time Series Forecasting for Forex with Deep Learning Techniques (2020)
- Predicting energy consumption using deep learning methods (2020)
- Deep learning framework to forecast electricity consumption of households and industries using long short-term memory network (2020)
- Predicting Weather Patterns Using LSTMs and Multiple Spatiotemporal Datasets (2020)
- Long Short Term Memory Networks for Anomaly Detection in Time Series (2015)
- A Review of Evaluation Metrics in Machine Learning Algorithms
0 of 30 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.