Application and Development of Quantitative Factor Mining in
Financial Markets
Tielong Ma
a
Deparment of Economics, University of Maryland College Park, College Park, Maryland, U.S.A.
Keywords: Quantitative Finance, Factor Mining, Financial Engineering, Market.
Abstract: Quantitative factor mining is a crucial element of modern quantitative investment, focusing on identifying
indicators that can predict asset price movements through data analysis. This paper systematically examines
five key factors influencing the effectiveness of quantitative factor mining: data quality, factor construction
methods, market environment, model selection, and computational resources, and provides targeted
optimization strategies. The research highlights that data quality serves as the foundation of factor mining,
while factor construction methods and model selection directly influence the predictive accuracy of factors.
Market environment shifts may cause factor degradation, and sufficient computational resources are essential
for efficient factor mining The paper explores traditional factor mining approaches, including fundamental
analysis, technical analysis, statistical analysis, and macroeconomic analysis, as well as prediction methods
based on major influencing factors, such as linear regression models, multi-factor models, time series models,
machine learning models, and ensemble learning techniques, summarizing the strengths and limitations of
each. Finally, the paper suggests future research directions, such as integrating multiple factor construction
methods and leveraging advanced machine learning techniques to enhance factor mining efficiency and
accuracy.
1 INTRODUCTION
Quantitative investment, as an important part of
modern finance, relies on data-driven decision-
making processes. With the rapid development of
information technology and the globalization of
financial markets, quantitative investment has
gradually shifted from traditional qualitative analysis
to quantitative analysis based on big data and
algorithms (Hull, 2018). Quantitative factor mining is
a core component of quantitative investment, aiming
to extract indicators that can predict asset price
movements through data analysis. These factors can
be fundamental factors based on company financial
data, technical factors based on market trading data,
or macroeconomic factors based on macroeconomic
data. The effectiveness of quantitative factor mining
directly impacts the returns and risks of investment
strategies, making it a critical issue in quantitative
research.
In recent years, the complexity and volume of
financial markets have exploded. High-frequency
a
https://orcid.org/0009-0002-7012-7093
trading, algorithmic trading, and the
interconnectedness of global markets have made
traditional investment analysis methods inadequate in
coping with increasingly complex market
environments. At the same time, the rapid
development of big data technology, cloud
computing, and artificial intelligence has provided
new tools and methods for quantitative investment.
As a core component of quantitative investment,
quantitative factor mining not only helps investors
identify potential opportunities in the market but also
explains cross-sectional differences in asset returns
through the construction of multi-factor models.
However, as market environments change and data
volumes increase, the difficulty of factor mining
continues to rise. Meanwhile, the effectiveness of
factor mining is influenced by multiple factors. First,
data quality is the foundation of factor mining. High-
quality data provides accurate information, while
low-quality data may lead to inaccurate factor
construction, thereby affecting the predictive power
of models. Second, the choice of factor construction
Ma, T.
Application and Development of Quantitative Factor Mining in Financial Markets.
DOI: 10.5220/0013688300004670
Paper published under CC license (CC BY-NC-ND 4.0)
In Proceedings of the 2nd International Conference on Data Science and Engineering (ICDSE 2025), pages 299-305
ISBN: 978-989-758-765-8
Proceedings Copyright © 2025 by SCITEPRESS – Science and Technology Publications, Lda.
299

methods directly impacts the predictive power of
factors. Traditional factor construction methods such
as fundamental analysis and technical analysis are
easy to understand but may perform poorly in
complex market environments. In recent years, with
the development of machine learning technology,
more researchers have begun to explore how to use
machine learning methods to automatically generate
factors to capture nonlinear relationships in the
market.
This paper first systematically analyzes the key
factors affecting the effectiveness of quantitative
factor mining from five aspects: data quality, factor
construction methods, market environment, model
selection, and computational resources, and proposes
corresponding optimization suggestions. Then, this
paper explores traditional factor mining methods,
such as fundamental analysis, technical analysis,
statistical analysis, and macroeconomic analysis, as
well as prediction methods based on mainstream
influencing factors, such as linear regression models,
multi-factor models, time series models, machine
learning models, and ensemble learning methods.
Finally, this paper summarizes the advantages and
disadvantages of different prediction methods and
proposes future research directions.
2 RESEARCH ON FACTORS
AFFECTING QUANTITATIVE
FACTOR MINING
Quantitative factor mining is a core component of
quantitative investment, and its effectiveness is
influenced by multiple factors (Tortoriello, 2012).
This paper systematically analyzes the key factors
affecting the effectiveness of quantitative factor
mining from five aspects: data quality, factor
construction methods, market environment, model
selection, and computational resources, and proposes
corresponding optimization suggestions. Research
shows that data quality is the foundation of factor
mining, factor construction methods and model
selection directly impact the predictive power of
factors, changes in the market environment may lead
to factor failure, and computational resources are the
hardware guarantee for efficient factor mining. By
optimizing these influencing factors, the
effectiveness of quantitative factor mining can be
significantly improved, thereby providing strong
support for the construction of quantitative
investment strategies. Extracting indicators that can
predict asset price movements through data analysis
directly impacts the returns and risks of investment
strategies. As the complexity and volume of financial
markets continue to increase, the role of quantitative
factor mining in investment decision-making
becomes increasingly important. However, the
effectiveness of factor mining is influenced by
multiple factors, including data quality, factor
construction methods, market environment, model
selection, and computational resources. This paper
aims to systematically analyze these influencing
factors, provide references for quantitative
researchers, and explore how to improve the
effectiveness of factor mining by optimizing these
factors.
2.1 Data Quality
Data is the foundation of quantitative factor mining,
and data quality directly affects the effectiveness of
factor mining. First, the completeness of data is
crucial. Missing data can lead to inaccurate factor
construction, thereby affecting the predictive power
of models. For example, when constructing financial
factors, if some companies' financial data is missing,
it may lead to poor performance of the factor in
backtesting. Second, the accuracy of data is also a key
issue. Incorrect data (such as outliers) can introduce
noise and reduce the effectiveness of factors. For
example, outliers in stock price data may lead to
errors in the calculation of technical indicators.
Additionally, the frequency and source of data also
affect the effectiveness of factor mining. High-
frequency data may contain more information but
may also introduce more noise. The quality of
different data sources (such as exchange data, third-
party data) may vary, so choosing reliable data
sources is an important step in improving data quality.
2.2 Factor Construction Methods
Factor construction methods are the core of
quantitative factor mining, and their choice directly
impacts the predictive power of factors. Traditional
factors, such as price-to-earnings ratio (PE), price-to-
book ratio (PB), etc., are easy to understand but may
be outdated. Technical indicators, such as moving
averages (MA), relative strength index (RSI), etc., are
suitable for short-term predictions. Fundamental
factors, such as revenue growth rate, net profit growth
rate, etc., are suitable for long-term investments.
Sentiment factors are extracted from news and social
media through natural language processing and are
suitable for event-driven strategies. In recent years,
machine learning-generated factors have gradually
ICDSE 2025 - The International Conference on Data Science and Engineering
300

become a research hotspot (López de Prado, 2018).
By using methods such as deep learning to
automatically generate factors, nonlinear
relationships may be captured, thereby improving the
predictive power of factors. However, machine
learning-generated factors also carry the risk of
overfitting and thus need to be used with caution.
2.3 Market Environment
Changes in the market environment can affect the
effectiveness of factors. First, market style shifts are
an important factor. For example, value factors may
fail in bull markets, while momentum factors may fail
in volatile markets. Second, changes in the
macroeconomic environment can also affect the
performance of factors. For example, changes in
interest rates, inflation, and other macroeconomic
factors may lead to the failure of certain factors.
Additionally, changes in the structure of market
participants are also a factor that cannot be ignored.
For example, as the proportion of institutional
investors increases, the effectiveness of certain
factors may decline. Therefore, in the process of
factor mining, it is necessary to fully consider
changes in the market environment and dynamically
adjust factors and models.
2.4 Model Selection
The selection and optimization of models are crucial
to the effectiveness of factor mining (Fabozzi et al.,
2010). Linear models, such as multiple linear
regression, are simple but may not capture nonlinear
relationships. Nonlinear models, such as random
forests and support vector machines, can capture
complex relationships but may overfit. Deep learning
models, such as neural networks, are suitable for
high-dimensional data but require significant
computational resources. Model ensemble methods,
such as stacking and boosting, can improve the
robustness of models. In practical applications, it is
necessary to select appropriate models based on
specific problems and data characteristics, and
optimize model parameters through methods such as
cross-validation to avoid overfitting.
2.5 Computational Resources
Computational resources are the hardware foundation
of quantitative factor mining. First, computing power
is a key factor affecting the efficiency of factor
mining. High-performance computing equipment can
accelerate the process of factor mining and
backtesting, thereby improving research efficiency.
Second, storage capacity is also an important issue.
Large-scale data requires sufficient storage space, so
it is necessary to reasonably plan storage resources.
Additionally, algorithm optimization is also an
important means to improve computational
efficiency. Efficient algorithms can reduce the
consumption of computational resources, thereby
achieving more efficient factor mining with limited
computational resources.
2.6 Optimization Suggestions
To improve the effectiveness of quantitative factor
mining, optimization can be carried out from the
following aspects. First, improve data quality.
Improve data completeness through data cleaning and
interpolation, and choose reliable data sources.
Second, diversify factor construction methods.
Combine traditional factors and machine learning-
generated factors to improve the predictive power of
factors. Third, dynamically adjust models.
Dynamically adjust factors and models based on
changes in the market environment. Finally, optimize
computational resources. Use distributed computing
and efficient algorithms to improve computational
efficiency.
2.7 Conclusion on Factor Mining
The effectiveness of quantitative factor mining is
influenced by multiple factors, including data quality,
factor construction methods, market environment,
model selection, and computational resources. By
systematically analyzing these influencing factors
and taking corresponding optimization measures, the
effectiveness of factor mining can be improved,
thereby enhancing the performance of quantitative
investment strategies. Future research can further
explore how to combine multiple factor construction
methods and how to use more advanced machine
learning techniques to improve the effectiveness of
factor mining.
3 PREDICTION METHODS
3.1 Research on Traditional Factor
Mining Methods
3.1.1 Fundamental Analysis
Fundamental analysis is one of the most traditional
methods of factor mining, mainly through the
Application and Development of Quantitative Factor Mining in Financial Markets
301

analysis of company financial data and operating
conditions to construct factors (Chincarini & Kim,
2006). Commonly used fundamental factors include
price-to-earnings ratio (PE), price-to-book ratio (PB),
dividend yield, revenue growth rate, and net profit
growth rate. These factors reflect a company's
profitability, growth potential, and valuation level,
and are suitable for long-term investment strategies.
For example, stocks with low price-to-earnings ratios
and high dividend yields are usually considered value
stocks with high investment value. However, the
limitation of fundamental analysis is its strong
reliance on financial data, which may be lagging and
easily affected by accounting policies. Additionally,
fundamental analysis usually assumes that the market
can effectively reflect a company's intrinsic value, but
in actual markets, this assumption may not hold.
3.1.2 Technical Analysis
Technical analysis is a method of predicting future
price trends by analyzing historical price and trading
volume data. Commonly used technical indicators
include moving averages (MA), relative strength
index (RSI), Bollinger Bands, and MACD (moving
average convergence divergence). These indicators
capture price trends, market sentiment, and
overbought or oversold conditions to construct
factors. For example, moving averages can be used to
identify trends, while the relative strength index can
be used to judge overbought or oversold conditions in
the market. The advantage of technical analysis is that
it is suitable for short-term trading strategies and
responds quickly to market changes. However, the
limitation of technical analysis is its strong reliance
on historical data and its susceptibility to market
noise. Additionally, technical analysis usually
assumes that historical price patterns will repeat, but
in actual markets, this assumption may not hold.
3.1.3 Statistical Analysis
Statistical analysis is a method of extracting factors
from data through mathematical and statistical
methods. Commonly used statistical methods include
principal component analysis (PCA), factor analysis,
and regression analysis. Principal component analysis
extracts the main features of data through
dimensionality reduction, factor analysis extracts
latent variables to explain the correlation between
observed variables, and regression analysis constructs
factors by establishing the relationship between
dependent and independent variables. For example,
regression analysis can be used to construct multi-
factor models, such as the Fama-French three-factor
model and the Carhart four-factor model (Fama &
French, 1993). The advantage of statistical analysis is
that it can handle high-dimensional data and extract
latent patterns. However, the limitation of statistical
analysis is its strong assumption about data
distribution and its susceptibility to multicollinearity
and overfitting. Additionally, statistical analysis
methods usually assume linear relationships between
variables, making it difficult to capture complex
nonlinear relationships.
3.1.4 Macroeconomic Analysis
Macroeconomic analysis is a method of constructing
factors by analyzing macroeconomic indicators.
Commonly used macroeconomic factors include
interest rates, inflation rates, GDP growth rates, and
unemployment rates. These factors reflect the impact
of the macroeconomic environment on asset prices
and are suitable for asset allocation and long-term
investment strategies. For example, a low-interest-
rate environment is usually favorable for the stock
market, while high inflation rates may lead to rising
bond yields. The advantage of macroeconomic
analysis is that it can capture the systemic impact of
the macroeconomic environment on the market.
However, the limitation of macroeconomic analysis
is its high requirement for data timeliness, and
changes in macroeconomic indicators are usually
slow, making it difficult to use for short-term trading
strategies. Additionally, macroeconomic analysis
usually assumes a stable relationship between
macroeconomic indicators and asset prices, but in
actual markets, this assumption may not hold.
3.1.5 Advantages and Disadvantages of
Traditional Methods
Traditional methods have the following advantages in
factor mining. First, traditional methods are usually
based on economic and financial theories, with clear
logic and easy understanding and interpretation.
Second, the computational complexity of traditional
methods is low, making them suitable for large-scale
data processing. Finally, many traditional factors
have shown stable performance in long-term
backtesting, with high reliability. However,
traditional methods also have the following
limitations: First, traditional methods have high
requirements for data completeness and accuracy, and
data quality issues may affect the effectiveness of
factor mining. Second, traditional methods usually
assume linear relationships between variables,
making it difficult to capture complex nonlinear
relationships. Finally, traditional methods may fail in
ICDSE 2025 - The International Conference on Data Science and Engineering
302

the face of market style shifts and structural changes,
with poor adaptability.
3.2 Prediction Methods Based on
Mainstream Influencing Factors
3.2.1 Linear Regression Models
Linear regression models are one of the most basic
prediction methods, predicting by establishing a
linear relationship between dependent and
independent variables. In linear regression models,
influencing factors are used as independent variables,
and asset prices or their returns are used as dependent
variables. For example, price-to-earnings ratio (PE),
price-to-book ratio (PB), and other fundamental
factors can be used as independent variables to
predict future stock returns. The advantage of linear
regression models is that they are simple, easy to
understand, computationally efficient, and the results
are easy to interpret. However, the limitation of linear
regression models is that they assume linear
relationships between variables, making it difficult to
capture complex nonlinear relationships.
Additionally, linear regression models are sensitive to
outliers and susceptible to multicollinearity.
3.2.2 Multi-Factor Models
Multi-factor models predict asset prices by
combining multiple influencing factors. Commonly
used multi-factor models include the Fama-French
three-factor model and the Carhart four-factor model.
The Fama-French three-factor model explains
differences in stock returns through market factors,
size factors, and value factors, while the Carhart four-
factor model adds momentum factors to the Fama-
French three-factor model (Carhart, 1997). The
advantage of multi-factor models is that they can
capture the combined effects of multiple influencing
factors and are suitable for explaining cross-sectional
differences in asset returns. However, the limitation
of multi-factor models is their strong reliance on
factor selection and weight allocation, and they make
it difficult to capture nonlinear relationships and
dynamic changes.
3.2.3 Time Series Models
Time series models predict future price trends by
analyzing time series data (Mengxia, 2023).
Commonly used time series models include
autoregressive models (AR), moving average models
(MA), autoregressive moving average models
(ARMA), and autoregressive integrated moving
average models (ARIMA). The advantage of time
series models is that they can capture trends and
periodicity in time series data and are suitable for
short-term predictions. However, the limitation of
time series models is their strong reliance on
historical data and difficulty in handling high-
dimensional data and external influencing factors.
Additionally, time series models usually assume that
time series are stationary, but in actual markets, this
assumption may not hold.
3.2.4 Machine Learning Models
Machine learning models automatically learn the
relationship between influencing factors and asset
prices through training data (Jansen, 2020).
Commonly used machine learning models include
decision trees, random forests, support vector
machines (SVM), and neural networks. The
advantage of machine learning models is that they can
capture complex nonlinear relationships and are
suitable for handling high-dimensional data. For
example, random forest models can be used to
combine multiple influencing factors to predict future
stock returns. However, the limitation of machine
learning models is their high requirement for data
volume and computational resources, and they are
susceptible to overfitting (McLean & Pontiff, 2016).
Additionally, the results of machine learning models
are usually difficult to interpret, which may affect
their application in investment decision-making.
3.2.5 Ensemble Learning Methods
Ensemble learning methods improve prediction
accuracy by combining the prediction results of
multiple models. Commonly used ensemble learning
methods include bagging, boosting, and stacking.
Bagging reduces variance by training multiple
models in parallel and averaging their results,
boosting reduces bias by training multiple models in
series and weighting their results, and stacking
constructs meta-models by combining the prediction
results of multiple models. The advantage of
ensemble learning methods is that they can improve
prediction accuracy and model robustness, making
them suitable for complex prediction tasks. However,
the limitation of ensemble learning methods is their
high computational complexity and strong
requirement for model diversity. Additionally, the
results of ensemble learning methods are usually
difficult to interpret, which may affect their
application in investment decision-making.
Application and Development of Quantitative Factor Mining in Financial Markets
303

3.2.6 Comparison of Different Prediction
Methods
Different prediction methods are suitable for different
market environments and data types. Linear
regression models are suitable for simple linear
relationships, multi-factor models are suitable for
explaining cross-sectional differences in asset
returns, time series models are suitable for short-term
predictions, machine learning models are suitable for
capturing complex nonlinear relationships, and
ensemble learning methods are suitable for improving
prediction accuracy and model robustness. In
practical applications, it is necessary to select
appropriate prediction methods based on specific
problems and data characteristics, and optimize
model parameters through methods such as cross-
validation to avoid overfitting.
4 CONCLUSIONS
The effectiveness of quantitative factor mining is
influenced by various factors, including data quality,
factor construction methods, market environment,
model selection, and computing resources. By
systematically analyzing these influencing factors
and adopting appropriate optimization measures, the
effectiveness of factor mining can be significantly
improved, thereby providing strong support for the
development of quantitative investment strategies.
This study shows that data quality is the foundation
of factor mining, factor construction methods and
model selection directly affect the predictive power
of factors, changes in the market environment may
lead to factor failure, and computing resources serve
as the hardware guarantee for efficient factor mining.
Traditional factor mining methods, fundamental
analysis, technical analysis, statistical analysis, and
macroeconomic analysis each have their own
strengths and weaknesses. Fundamental analysis is
suitable for long-term investment strategies but relies
heavily on financial data; technical analysis is
suitable for short-term trading strategies but is highly
dependent on historical data; statistical analysis can
handle high-dimensional data but relies heavily on
assumptions about data distribution; and
macroeconomic analysis can capture the systemic
impact of the macro environment on the market but
requires high timeliness of data.
For predictive methods based on mainstream
influencing factors, linear regression models, multi-
factor models, time series models, machine learning
models, and ensemble learning methods each have
their applicable scenarios. Linear regression models
are simple and easy to understand but struggle to
capture complex nonlinear relationships; multi-factor
models can capture the combined effects of multiple
influencing factors but rely heavily on factor selection
and weight allocation; time series models are suitable
for short-term forecasting but depend heavily on
historical data; machine learning models can capture
complex nonlinear relationships but require large
amounts of data and computational resources (Gu et
al., 2020); and ensemble learning methods can
improve predictive accuracy and model robustness
but have high computational complexity.
Future research could further explore how to
combine multiple factor construction methods and
how to leverage more advanced machine learning
techniques to improve the effectiveness of factor
mining (Binhui, 2024). Additionally, as computing
resources continue to advance, how to use them more
efficiently for factor mining is also a worthwhile
research direction. By continuously optimizing each
aspect of factor mining, the performance of
quantitative investment strategies will be further
enhanced.
REFERENCES
Binhui, W. 2024. Research on LSTM stock price prediction
model based on financial comments and functional
data. CNKI.
Carhart, M. M. 1997. On persistence in mutual fund
performance. Journal of Finance, 52(1), 57–82.
Chincarini, L., & Kim, D. 2006. Quantitative equity
portfolio management: An active approach to portfolio
construction and management. McGraw-Hill
Education.
Fama, E. F., & French, K. R. 1993. Common risk factors in
the returns on stocks and bonds. Journal of Financial
Economics, 33(1), 3–56.
Gu, S., Kelly, B., & Xiu, D. 2020. Empirical asset pricing
via machine learning. Review of Financial Studies,
33(5), 2223–2273.
Hull, J. C. 2018. Financial data analysis (2nd ed.). Pearson
Education.
Jansen, S. 2020. Machine learning for algorithmic trading:
Predictive models to extract insights from data and
build trading strategies (2nd ed.). O'Reilly Media.
López de Prado, M. 2018. Advances in financial machine
learning. John Wiley & Sons.
McLean, R. D., & Pontiff, J. 2016. Does academic research
destroy stock return predictability? Journal of Finance,
71(1), 5–32.
Mengxia, L. 2023. Pattern mining and correlation analysis
methods for financial time series. CNKI.
ICDSE 2025 - The International Conference on Data Science and Engineering
304

Tortoriello, R. 2012. Quantitative strategies for achieving
alpha: The standard and poor's approach to testing your
investment choices. McGraw-Hill Education.
Application and Development of Quantitative Factor Mining in Financial Markets
305