Paper: arXiv 2601.08896
Authors: Sahaj Raj Malla, Shreeyash Kayastha, Rumi Suwal, Harish Chandra Bhandari, Rajendra Adhikari
Abstract
This study develops a robust machine learning framework for one-step-ahead forecasting of daily log-returns in the Nepal Stock Exchange (NEPSE) Index using the XGBoost regressor. A comprehensive feature set is engineered, including lagged log-returns (up to 30 days) and established technical indicators such as short- and medium-term rolling volatility measures and the 14-period Relative Strength Index. Hyperparameter optimization is performed using Optuna with time-series cross-validation on the initial training segment. Out-of-sample performance is rigorously assessed via walk-forward validation under both expanding and fixed-length rolling window schemes across multiple lag configurations, simulating real-world deployment and avoiding lookahead bias. Predictive accuracy is evaluated using root mean squared error, mean absolute error, coefficient of determination (R-squared), and directional accuracy on both log-returns and reconstructed closing prices. Empirical results show that the optimal configuration, an expanding window with 20 lags, outperforms tuned ARIMA and Ridge regression benchmarks, achieving the lowest log-return RMSE (0.013450) and MAE (0.009814) alongside a directional accuracy of 65.15%. While the R-squared remains modest, consistent with the noisy nature of financial returns, primary emphasis is placed on relative error reduction and directional prediction. Feature importance analysis and visual inspection further enhance interpretability. These findings demonstrate the effectiveness of gradient boosting ensembles in modeling nonlinear dynamics in volatile emerging market time series and establish a reproducible benchmark for NEPSE Index forecasting.
Complexity vs Empirical Score
- Math Complexity: 4.0/10
- Empirical Rigor: 8.5/10
- Quadrant: Street Traders — empirical and applied, lower math complexity
Why this score: The paper uses standard ML techniques (XGBoost, ARIMA, Ridge) with accessible feature engineering, showing low-to-moderate mathematical density. However, it demonstrates high empirical rigor through rigorous walk-forward validation, hyperparameter optimization (Optuna), and detailed backtest-ready metrics (RMSE, MAE, directional accuracy) on real historical data.
Research Flowchart
flowchart TD
A["Research Goal: Develop a robust ML framework for one-step-ahead forecasting of NEPSE daily log-returns"] --> B["Data & Feature Engineering"]
B --> C["Model Optimization"]
C --> D["Validation Strategy"]
D --> E["Key Outcomes"]
subgraph B ["Data & Feature Engineering"]
B1["NEPSE Index Historical Data"]
B2["Engineered Features: Lagged returns (up to 30 days),<br>Rolling Volatility, RSI (14-period)"]
end
subgraph C ["Model Optimization"]
C1["XGBoost Regressor"]
C2["Hyperparameter Tuning via Optuna<br>using Time-Series CV"]
end
subgraph D ["Validation Strategy"]
D1["Walk-Forward Validation<br>Expanding & Rolling Windows"]
D2["Benchmarks: ARIMA & Ridge Regression"]
end
subgraph E ["Key Outcomes"]
E1["Optimal Config: Expanding Window, 20 Lags"]
E2["Superior Performance vs. Benchmarks<br>Lowest RMSE (0.013450) & MAE (0.009814)"]
E3["65.15% Directional Accuracy"]
E4["Feature Importance Analysis for Interpretability"]
end
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 robust XGBoost-based machine learning framework for one-step-ahead forecasting of daily NEPSE Index log‑returns.
Source sentence
This study develops a robust machine learning framework for one-step-ahead forecasting of daily log-returns in the Nepal Stock Exchange (NEPSE) Index using the XGBoost regressor.
Assumes 5
- Log‑returns are stationary and time‑additive, justifying their use as the target variable. statistical
Source sentence
Log-returns are employed due to their desirable statistical properties, including approximate normality and time-additivity, which facilitate modeling in financial time
- The model relies exclusively on price‑derived features and does not incorporate exogenous macro or news variables. data
Source sentence
The framework relies solely on price-derived features and does not incorporate exogenous variables, such as macroeconomic indicators or news sentiment, which could further improve explanatory power.
- Walk‑forward validation with expanding or fixed‑length windows prevents look‑ahead bias and mimics real‑world deployment. model
Source sentence
Out-of-sample performance is rigorously assessed via walk-forward validation under both expanding and fixed-length rolling window schemes across multiple lag configurations, simulating real-world deployment and avoiding lookahead bias.
- Hyperparameter tuning is performed only on the initial training segment to avoid data leakage. data
Source sentence
Hyperparameter optimization is performed using Optuna with time-series cross-validation on the initial training segment.
Show 1 more
- The final dataset is split chronologically, reserving the last 20% for out‑of‑sample testing to ensure realistic forecasting conditions. data
Source sentence
The final dataset is split chronologically, reserving the last 20% of observations exclusively for out-of-sample testing to avoid data leakage and ensure realistic forecasting conditions.
Data
NEPSE Index · equities · daily
- NEPSE Index daily historical records market-prices public
Source sentence
The dataset comprises daily historical records of the NEPSE Index, sourced from the third-party portal NepseAlpha
Costs
Not applicable
Finds 3
- The optimal expanding‑window XGBoost configuration (20 lags) achieves the lowest log‑return RMSE of 0.013450, MAE of 0.009814, and directional accuracy of 65.15%, outperforming tuned ARIMA and Ridge benchmarks. performance
Source sentence
The optimal configuration—an expanding window with 20 lags—outperforms tuned ARIMA and Ridge regression benchmarks, achieving the lowest log-return RMSE (0.013450) and MAE (0.009814) alongside a directional accuracy of 65.15%.
- Diebold‑Mariano tests indicate the XGBoost model’s predictive superiority over Ridge and ARIMA (p‑values < 0.01). statistical
Source sentence
The DM statistics indicate significant superiority of the optimal XGBoost model (p-values < 0.01 in both pairwise comparisons).
- Feature importance analysis provides interpretable insights into momentum, mean reversion, and volatility clustering. other
Source sentence
Feature importance analysis and visual inspection further enhance interpretability.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 28 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 (21)
- Volatility of daily Nepal stock exchange (NEPSE) index return: A GARCH family models (2021)
- Predicting NEPSE Index: An ARIMA-Based Model (2024)
- Financial Time Series Forecasting with Deep Learning : A Systematic Literature Review: 2005-2019 (2019) — link
- Predicting Nepse Index Price Using Deep Learning Models (2022) — link
- Stock price forecasting with deep learning: A comparative study (2020)
- Predicting the direction of NEPSE index movement with news headlines using machine learning (2024)
- A note on the validity of cross-validation for evaluating autoregressive time series prediction (2018)
- Deep learning models for price forecasting of financial time series: A review of recent advancements: 2020–2022 — → 2305.04811
and 13 more
- Machine learning techniques and data for stock market forecasting: A literature review (2022)
- Nepal stock exchange prediction using support vector regression and neural networks (2018)
- Analysis of financial time series (2005)
- Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining (2016)
- Optuna: A next-generation hyperparameter optimization framework (2019)
- Forecasting: principles and practice (2018)
- Nepal stock market movement prediction with machine learning (2021)
- Combining principal component analysis, discrete wavelet transform and XGBoost to trade in the financial markets (2019)
- Forecasting method of stock market volatility in time series data based on mixed model of ARIMA and XGBoost (2020)
- Forecasting of forex time series data based on deep learning (2019)
- A hybrid model integrating deep learning with investor sentiment analysis for stock price prediction (2021)
- Forecasting gold price with the XGBoost algorithm and SHAP interaction values (2024)
- Evaluating XGBoost Algorithm for Nepali Stock Price Prediction (2024)
1 of 21 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.