assumption of parameter stability over time does not
hold well in the volatile stock market, reducing
predictive accuracy. The LSTM model, while
designed to handle long-term dependencies,
struggled with capturing equity premium trends due
to its high data and hyperparameter tuning
requirements, and its sensitivity to sequence length.
The bagged trees model, on the other hand, benefits
from capturing complex interactions and non-linear
relationships through ensemble learning, reducing
over-fitting and variance. However, it exhibited
frequent abrupt fluctuations, indicating noise
sensitivity and potential over-fitting to training data,
along with insufficient temporal adaptation,
affecting its long-term prediction reliability. These
differences indicated by its negative R-squared
values, which highlight the inherent unpredictability
of the equity premium.
3.3 Discussion
The unpredictability of the equity premium is due to
several factors: limited access to comprehensive
information in real-world markets, the inherent noise
and complexity of financial markets exhibiting non-
stationary and chaotic behaviors, and the simplicity
of the models used in the analysis, which lack the
sophisticated optimization of Genetic Algorithms-
enhanced (GAs-enhanced) models. Additionally, the
tendency of complex models to over-fit historical
data can result in poor generalization in OOS
predictions.
The current implementation is not without areas
to improve. First, incorporating alternative data
sources, such as sentiment from social media or
news, could provide more insights and improve
predictions. Second, advanced noise-filtering
techniques like wavelet transforms or Kalman filters
could help isolate useful patterns from market noise.
Third, using hybrid models and advanced
optimization techniques like GAs may enhance
model performance. Moreover, future research
should also focus on preventing over-fitting by
applying better regularization methods and using
robust validation techniques. Developing dynamic
models that adapt to changing market conditions,
such as those using reinforcement learning, could
also lead to more accurate predictions.
4 CONCLUSIONS
This study reevaluated stock price prediction by
comparing the performance of traditional and
advanced machine learning techniques. A rolling
linear regression model and advanced models such
as LSTM and bagged trees were used to predict
equity premiums, evaluating their performance using
OOS-R
2
, RMSE, and MAE. This study concludes
that while advanced machine learning models offer
improved performance metrics, the equity premium
remains fundamentally challenging to predict
accurately. Contributing factors to this
unpredictability include limited access to
comprehensive information, the inherent noise and
complexity of financial markets, the simplicity of
models used, and the over-fitting tendency of
complex models on historical data. To improve
future predictions, incorporating alternative data
sources, advanced noise-filtering techniques, hybrid
models, and better regularization methods, as well as
developing dynamic models that adapt to changing
market conditions, are recommended.
REFERENCES
Ang, A., & Bekaert, G. 2007. Stock return predictability:
Is it there? The Review of Financial Studies, 20(3),
651-707.
Campbell, J. Y., & Shiller, R. J. 1988. The dividend-price
ratio and expectations of future dividends and discount
factors. The review of financial studies, 1(3), 195-228.
Chen, Y., Cheng, Q., Cheng, Y., Yang, H., & Yu, H. 2018.
Applications of recurrent neural networks in
environmental factor forecasting: A review. Neural
Computation, 30(11), 2855-2881.
Fama, E. F., & French, K. R. 1996. Multifactor
explanations of asset pricing anomalies. The Journal
of Finance, 51(1), 55-84.
Gu, S., Kelly, B., & Xiu, D. 2020. Empirical asset pricing
via machine learning. The Review of Financial Studies,
33(5), 2223-2273.
Ingolfsson, T. M. 2021. Insights into LSTM architecture.
Retrieved from https://thorirmar.com/post/insight_into_
lstm/
Lettau, M., & Ludvigson, S. 2001. Consumption,
aggregate wealth, and expected stock returns. the
Journal of Finance, 56(3), 815-849.
Mehra, R., & Prescott, E. C. 2003. The equity premium in
retrospect. Handbook of the Economics of Finance, 1,
889-938.
Opitz, D., & Maclin, R. 1999. Popular ensemble methods:
An empirical study. Journal of Artificial Intelligence
Research, 11, 169-198.
Welch, I., & Goyal, A. 2008. A comprehensive look at the
empirical performance of equity premium prediction.
The Review of Financial Studies, 21(4), 1455-1508.
Wu, Y., Ke, Y., Chen, Z., Liang, S., Zhao, H., & Hong, H.
2020. Application of alternating decision tree with
AdaBoost and bagging ensembles for landslide
susceptibility mapping. Catena, 187, 104396.