Paper: arXiv 2502.08144
Authors: Peiwan Wang, Chenhao Cui, Yong Li
Abstract
In recent years, the dominance of machine learning in stock market forecasting has been evident. While these models have shown decreasing prediction errors, their robustness across different datasets has been a concern. A successful stock market prediction model minimizes prediction errors and showcases robustness across various data sets, indicating superior forecasting performance. This study introduces a novel multiple lag order probabilistic model based on trend encoding (TeMoP) that enhances stock market predictions through a probabilistic approach. Results across different stock indexes from nine countries demonstrate that the TeMoP outperforms the state-of-the-art machine learning models in predicting accuracy and stabilization.
Complexity vs Empirical Score
- Math Complexity: 5.5/10
- Empirical Rigor: 6.5/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper presents a probabilistic model with multiple lag orders and trend encoding, involving statistical and fuzzy logic concepts, indicating moderate math complexity. Empirically, it tests the model across nine international stock indices and compares it to state-of-the-art ML models, showing a data-heavy and backtest-ready approach.
Research Flowchart
flowchart TD A["Research Goal: Non-ML Approach for Robust Stock Forecasting"] --> B["Methodology: Trend-encoded Multi-order Probabilistic Model"] B --> C["Input: Global Stock Index Data<br>9 Countries"] C --> D["Computation: Trend Encoding &<br>Multi-order Probabilistic Analysis"] D --> E["Outcomes: Superior Accuracy &<br>Stabilized Forecasts vs. SOTA ML"]
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 introduces a novel multiple lag order probabilistic model based on trend encoding (TeMoP) for stock market forecasting.
Source sentence
This study introduces a novel multiple lag order probabilistic model based on trend encoding (TeMoP) that enhances stock market predictions through a probabilistic approach.
Assumes 2
- The time series data are non‑stationary, so models trained on a fixed lag order may lack robustness. data
Source sentence
In the real world, time series data often exhibit non-stationary properties. Therefore, models trained based on samples with a certain lag order tend to exhibit poor robustness.
- Trend features of the samples significantly impact forecasting performance. model
Source sentence
Hypothesis: The trend features of the sample significantly impact the performance of time series trend forecasting.
Data
- Nine major global financial market indices (daily closing prices) market-prices public
Source sentence
There are nine major global financial market indices as data sets to assess models. In the experiment, the daily closing price of each index is selected for analysis.
Costs
Unclear
Finds 2
- TeMoP achieves the highest mean accuracy (ACC) across all nine indices and the smallest standard deviation, indicating the smallest prediction error and good robustness. performance
Source sentence
The mean ACC of TeMoP takes the largest value among all models. The standard deviation of ACC of TeMoP, LGBM, and LR is smaller compared to other models. Therefore, TeMoP exhibits the smallest prediction error and good robustness across data sets when measuring the prediction performance of the models based on the metr
- TeMoP obtains the largest F1 scores on most data sets, matching or closely approaching the best F1 values, demonstrating strong predictive accuracy. predictive-accuracy
Source sentence
In Figure 4 and Table 6, TeMoP has the largest F1 on the rest of the data sets except for data set Nik. For data set Nik, compared to the largest F1 0.674, F1 of TeMoP is 0.637, a difference of 0.1%.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 25 related papers). Coverage is partial, so this is neither replication nor refutation.
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.
Cited References (48)
- Deep learning with multiple scale attention and direction regularization for asset price prediction (2021) — link
- Machine learning advances for time series forecasting (2023)
- McVCsB: A new hybrid deep learning network for stock index prediction (2023) — link
- Traffic flow prediction with big data: A deep learning approach (2014)
- Deep learning based ensemble approach for probabilistic wind power forecasting (2017)
- An ARIMA-LSTM model for predicting volatile agricultural price series with random forest technique Image 1 (2023) — link
- Recent advances in decision trees: An updated survey (2023)
- Research on stock trend prediction method based on optimized random forest (2023)
and 40 more
- Adaptive Temporal-Frequency Network for Time-Series Forecasting (2022) — link
- Combining embeddings and fuzzy time series for high-dimensional time series forecasting in internet of energy applications (2023) — link
- Self-paced ARIMA for robust time series prediction (2023) — link
- A Time-Series Feature-Based Recursive Classification Model to Optimize Treatment Strategies for Improving Outcomes and Resource Allocations of COVID-19 Patients (2022) — link
- Distributed ARIMA models for ultra-long time series (2023) — link
- LRF: A logically randomized forest algorithm for classification and regression problems (2023) — link
- An improved random forest based on the classification accuracy and correlation measurement of decision trees (2024)
- Random subsequence forests (2024) — link
- Adversarial Joint-Learning Recurrent Neural Network for Incomplete Time Series Classification (2022) — link
- Stock Movement Prediction Based on Bi-Typed Hybrid-Relational Market Knowledge Graph via Dual Attention Networks (2023) — link
- Traffic Speed Prediction Based on Time Classification in Combination With Spatial Graph Convolutional Network (2023) — link
- A Fuzzy-Probabilistic Representation Learning Method for Time Series Classification (2024) — link
- Distributed evolutionary hyperparameter optimization for fuzzy time series (2020)
- Bayesian network based probabilistic weighted high-order fuzzy time series forecasting (2024)
- Probabilistic forecasting with fuzzy time series (2019)
- Short-term load forecasting by using a combined method of convolutional neural networks and fuzzy time series (2019)
- Evolving fuzzy time series for spatio-temporal forecasting in renewable energy systems (2021)
- Combining embeddings and fuzzy time series for high-dimensional time series forecasting in internet of energy applications (2023)
- pyFTS: fuzzy time series for python.(2018)
- Modeling and predicting rainfall time series using seasonal-trend decomposition and machine learning (2022)
- Forecasting gold price using a novel hybrid model with ICEEMDAN and LSTM-CNN-CBAM (2022)
- A seasonal-trend decomposition-based dendritic neuron model for financial time series prediction (2021) — link
- A novel hybrid STL-transformer-ARIMA architecture for aviation failure events prediction (2024)
- A weighted EMD-based prediction model based on TOPSIS and feed forward neural network for noised time series (2017) — link
- A SVR–ANN combined model based on ensemble EMD for rainfall prediction (2018) — link
- Arctic short-term wind speed forecasting based on CNN-LSTM model with CEEMDAN (2024) — link
- Short-term load forecasting based on CEEMDAN and dendritic deep learning (2024) — link
- Multi-step-ahead stock price prediction using recurrent fuzzy neural network and variational mode decomposition — → 2212.14687
- Oil price forecasting: A hybrid GRU neural network based on decomposition–reconstruction methods (2023) — link
- Stock price forecasting based on LLE-BP neural network model (2020)
- Day-ahead spatiotemporal solar irradiation forecasting using frequency-based hybrid principal component analysis and neural network (2019) — link
- A distributed algorithm for scalable fuzzy time series (2019)
- Forecasting monthly gas field production based on the CNN-LSTM model (2022) — link
- Gold volatility prediction using a CNN-LSTM approach (2020) — link
- Hesitant fuzzy time series forecasting model of higher order based on one and two-factor aggregate logical relationship (2023) — link
- A fuzzy time series forecasting model with both accuracy and interpretability is used to forecast wind power (2024) — link
- Inceptiontime: Finding alexnet for time series classification (2020)
- Integration of residual network and convolutional neural network along with various activation functions and global pooling for time series classification (2019)
- Stock movement prediction based on bi-typed hybrid-relational market knowledge graph via dual attention networks (2022)
- STD: a seasonal-trend-dispersion decomposition of time series (2023)
1 of 48 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.