Multi-Dimensional Analysis and Exploration of Asset Pricing

Yujia Hao

Economic and Finance, University of Melbourne, Parkville Victoria, 3010, Australia

Keywords: Asset Pricing, Influencing Factors, Pricing Methods.

Abstract: Asset pricing is a core issue in the financial market and is affected by various external and internal factors.

This study explores the role of policy changes, socio-economic uncertainty, and corporate internal governance

structure in asset pricing. It analyzes the advantages and disadvantages of traditional pricing methods (such

as CAPM, APT, and multi-factor models) and emerging data-driven methods (such as machine learning and

deep learning). Policy factors mainly affect market liquidity, capital costs, and investor expectations through

monetary policy, fiscal policy, and financial regulation, which affect asset prices. Socio-economic uncertainty,

such as economic crises, wars, and natural disasters, can exacerbate market volatility and affect asset

valuations. A company's profitability, capital structure, and governance level determine its market value. In

addition, the development of data science, machine learning, and deep learning in asset pricing continues to

improve prediction accuracy, but it still faces challenges such as black box problems, overfitting, and data

reliability. In the future, combining behavioral finance, data-driven methods, and research on explainable

artificial intelligence (XAI) is expected to improve the accuracy and applicability of asset pricing models and

promote the transformation of financial analysis into intelligence.

1 INTRODUCTION

Asset pricing is a core issue in the financial market

that directly affects investment decisions, corporate

financing, and market stability. In theory, asset prices

should reflect the discounted value of their future

cash flows, but the actual market is affected by

multiple factors, such as policies, the economic

environment, and investor behavior, resulting in price

fluctuations. Understanding these factors and

establishing a reasonable pricing model is crucial for

optimizing investment strategies and financial

supervision.

Traditional pricing methods, such as the capital

asset pricing model (CAPM), arbitrage pricing theory

(APT), and the Fama-French multi-factor model,

emphasize the decisive role of market risk premium

on asset returns. However, these models have

limitations in explaining market anomalies and

nonlinear problems. In addition, behavioral finance

shows that irrational investor behaviors, such as

herding and loss aversion, can affect asset pricing and

cause the market to deviate from theoretical

predictions.

https://orcid.org/0009-0003-6852-9234

With the rapid development of big data

technology and artificial intelligence, asset pricing

research has entered a new stage in recent years. Data-

driven methods such as machine learning (ML) and

deep learning (DL) have become a new trend in asset

pricing research. These methods can capture complex

nonlinear relationships and improve prediction

accuracy. For example, neural networks outperform

traditional factor models in predicting stock returns,

and natural language processing (NLP) technology

also makes market sentiment analysis possible.

However, these methods still face challenges such as

black box problems, overfitting, and data quality,

which, to some extent, limit their widespread

application in the financial field.

This study will analyze the impact of policies,

socio-economic environment, and corporate internal

governance on asset pricing and compare the

advantages and disadvantages of traditional and

machine learning methods. At the same time, it

explores how new technologies such as behavioral

finance, causal inference, and explainable artificial

intelligence (XAI) can optimize pricing models. This

study provides suggestions and inspiration for the

Hao, Y.

Multi-Dimensional Analysis and Exploration of Asset Pricing.

DOI: 10.5220/0013686000004670

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 253-258

ISBN: 978-989-758-765-8

253

direction and path of future asset pricing research

from a multi-dimensional and interdisciplinary

perspective by exploring the relevant influencing

factors and practical progress in asset pricing.

2 INFLUENCING FACTORS

Asset pricing is affected by various external and

internal factors, which not only affect the formation

of market prices but also determine investors' risk

preferences and market behavior. Policy changes,

economic and social uncertainties, and corporate

internal governance structures often affect asset price

fluctuations. The following will be analyzed based on

policy changes, social and economic uncertainties,

and internal corporate factors.

2.1 Policy

Government policies, mainly monetary, fiscal, and

financial regulatory policies, are essential factors

affecting asset pricing. These policies change the

assets' expected return and investment risk by

affecting market liquidity, capital costs, and corporate

profitability.

2.1.1 Monetary Policy

Monetary policy impacts asset pricing mainly by

affecting market interest rates, liquidity, and investor

expectations. The central bank controls inflation and

economic growth by adjusting benchmark interest

rates (such as the Federal Reserve's federal funds rate)

and open market operations, thereby affecting the

market's risk premium, discount rate, and capital cost,

and thus determining changes in asset prices

(Bernanke & Kuttner, 2005). For example, the Fed's

interest rate hike in 2022 raised the discount rate,

depressed asset prices, and pushed up corporate

financing costs, weakening profitability and reducing

stock valuations. At the same time, market risk

premiums rose, investors' risk aversion increased, and

funds flowed out of the stock market and turned to

bonds, leading to a contraction in market liquidity and

exacerbating downward pressure on the stock market

(Li, 2023).

2.1.2 Fiscal Policy

Fiscal policy (such as government spending and tax

adjustments) changes asset pricing by affecting

economic growth, corporate profits, and market

confidence. Expansionary policies (increasing

investment or reducing taxes) increase corporate cash

flow and profit expectations, reduce risk premiums,

and push up stock market valuations. However,

excessive fiscal deficits may lead to rising inflation

and increased government debt, increase discount

rates, and weaken market confidence, thereby

suppressing asset price increases (Agnello & Sousa,

2011).

2.1.3 Regulatory Policy

Changes in financial regulatory policies also affect

asset pricing. After the 2008 financial crisis, countries

strengthened financial market supervision, requiring

banks to increase capital adequacy ratios and limit

leverage operations, which reduced the volatility of

financial asset prices. For example, Basel III requires

banks to increase their capital adequacy ratio and

liquidity coverage ratio and restrict high-leverage

businesses, which increases banks' financing costs

and thus affects financial asset prices (Gratton, 2024).

In contrast, if the government relaxes regulation, such

as lowering the IPO threshold in the capital market, it

may increase investment opportunities in the market

and improve stock valuations (Gallimberti, Lambert,

& Xiao, 2021).

2.2 Socio-Economic

Socio-economic uncertainties, such as financial

crises, wars, and natural disasters, can significantly

affect market stability and reduce the predictive

power of asset pricing models. For example, during

the 2008 financial crisis, investors panicked, sold

high-risk assets, and turned to safe-haven assets such

as gold and government bonds, causing asset prices

to fluctuate sharply, showing investors' risk aversion

(Muir, 2017).

Wars and geopolitical conflicts, such as the 2022

Russia-Ukraine conflict, have led to increased risk

aversion in the market, affecting global stock and

commodity markets (Souza, 2020). In addition,

natural disasters, such as the COVID-19 pandemic,

have profoundly impacted global supply chains and

market liquidity, forcing companies to lower their

earnings expectations and thereby reducing stock

market valuations (Berkman & Malloch, 2021).

2.3 Internal Factors of the Enterprise

An enterprise's financial status, management

decisions, and corporate governance structure are the

core internal factors that affect asset pricing.

Profitability, capital structure, and cash flow status

ICDSE 2025 - The International Conference on Data Science and Engineering

254

determine the market's expectations of a company's

future earnings, affecting its valuation. Management's

strategic decisions, including mergers and

acquisitions, stock repurchases, and capital

allocation, directly affect the market's judgment of the

company's growth potential and financial soundness.

Overall, a sound financial position, reasonable

management decisions, and sound corporate

governance can improve asset valuations, while

financial instability, strategic errors, or governance

deficiencies may lead to a decline in asset prices

(Affes & Jarboui, 2023).

3 PRICING METHODS

3.1 Traditional Methods

3.1.1 CAPM Model

The capital asset pricing model (CAPM) is a core tool

in modern financial market price theory that is widely

used in investment decisions and corporate finance.

The Model was developed by William Sharpe, John

Lintner, and others in the 1960s based on modern

portfolio theory (Kenton, 2024). Its core formula is:

(



)



+β



[

(



)

−r



]

(

)

Among them, E(r



) is the expected return on the

asset, r



is the risk-free rate, β



represents the

systematic risk of the asset to the market portfolio,

and E(r



) is the expected return of the market

portfolio. CAPM believes that the market risk

premium is the main factor determining asset returns.

However, Fama and French pointed out that

CAPM has difficulty explaining the excess returns of

small-cap stocks and the high returns of high book-

to-market ratio stocks (Hayes, 2024a). In addition,

CAPM assumes that the market is entirely rational,

while behavioral finance research shows that

investors' irrational behavior can affect asset prices.

3.1.2 APT Model

The Arbitrage Pricing Theory (APT) is a multi-factor

asset pricing model that assumes that the linear

relationship between expected returns and a series of

macroeconomic factors can predict asset returns. The

Model helps value investing and identifies securities

that may be mispriced (Hayes, 2020). Its formula is:

E(r



)=R



+β







#(2)

Where E(r



) is the expected return of asset i, R



the risk-free rate, β



is the sensitivity of asset i to the

jth risk factor, and F



represents the number of risk

factors selected by the jth systematic risk factor (such

as GDP growth rate, interest rate changes).

APT is more flexible than CAPM and does not

require market equilibrium, but its main limitation is

that the theory does not propose relevant factors for

specific stocks or assets. A stock may be more

sensitive to one factor than another, and investors

must be able to perceive the source and sensitivity of

risk.

3.1.3 Multi-factor Model

The Fama-French three-factor model is an asset

pricing model based on the CAPM expansion in 1992.

Based on market risk factors, the market value factor

(SMB) and the book-to-market factor (HML) were

introduced to more accurately explain asset returns

(Hayes, 2024a).

E(r



)=r



+β



[E(r



)−r



]−β



SMB

+β



HML#(3)

The Model was later expanded to a five-factor

model, adding profitability and investment factors.

Compared with CAPM and APT, although it has

improved the ability to explain cross-sectional

returns, it still has defects. Its factors may be

redundant and multicollinear, the theoretical basis is

weak, and it fails to fully explain small-cap stocks'

excess returns and momentum effects. In addition, the

Model depends on the market environment, and its

effectiveness is unstable in different markets or

economic cycles, and it still cannot eliminate

abnormal phenomena. Despite this, it is still an

important tool for asset pricing and investment

analysis.

3.2 Machine Learning Methods

With today's financial markets becoming increasingly

complex and data availability improving, the

limitations of traditional linear asset pricing models

are becoming increasingly apparent. Machine

learning methods use computing power and data-

driven optimization to be more adaptable and

accurate in asset pricing and risk prediction. They can

automatically identify nonlinear relationships and

adapt to market dynamics, especially in high-

dimensional data environments.

Multi-Dimensional Analysis and Exploration of Asset Pricing

255

3.2.1 The Role of Machine Learning in Asset

Pricing

Machine learning introduces several key features into

asset pricing models, including processing large

amounts of data (such as historical financial market

data and corporate financial data) and using machine

learning models (such as deep learning, support

vector machines, etc.) to predict future price trends.

Applying machine learning (ML) models in asset

pricing has significantly improved prediction

accuracy, surpassing traditional linear models.

Advanced machine learning models (such as neural

networks, support vector machines (SVMs), and

gradient boosting trees (GBTs)) have higher R2

values and lower error metrics (such as root mean

square error RMSE and mean absolute error MAE)

(Fang & Taylor, 2021). These models achieve more

accurate predictions by capturing the complex

nonlinear relationships between asset prices and

multiple factors.

For example, machine learning models

significantly outperform traditional linear factor

models when predicting abnormal returns of

portfolios and show higher prediction accuracy. This

improvement is reflected in the optimization of

statistical indicators and the model's more profound

understanding and adaptability to market dynamics.

In the return prediction of China's A-share market, the

neural network model increased the R2 value from

0.31 to 0.46, far exceeding the traditional linear

model. This high prediction accuracy enhances the

reliability of investment decisions and provides

investors with better risk-adjusted returns (Li, 2018).

Although machine learning (ML) models perform

well in prediction accuracy, the "black box"

characteristics raise ethical issues about transparency

and accountability. Especially in the financial

industry, investors and regulators need to have a clear

understanding and interpretability of the model's

decision logic, but the high complexity of machine

learning models may lead to insufficient transparency

(Chen, Pelger, & Zhu, 2024).

3.2.2 The Role of Deep Learning in Asset

Pricing

Deep learning, especially neural network models, can

handle more complex market relationships and

overcome the limitations of traditional methods in

processing nonlinear and high-dimensional data.

The long-short portfolio strategy built based on

neural networks outperforms traditional linear

models in terms of annualized returns and Sharpe

ratio. Studies have shown that the Sharpe ratio of the

neural network model reaches 2.97, while that of the

linear model is only 1.93. This indicates that neural

networks can improve portfolio returns and

effectively reduce risks. Its significant performance

improvement stems from the fact that neural networks

can capture and utilize complex nonlinear

relationships and dynamic changes in asset pricing,

thereby providing more accurate and reliable

investment signals (Pan, 2019).

Unlike traditional financial models, neural

networks' black-box nature may face regulatory

barriers in practical applications. In addition, deep

learning relies on a large amount of data training.

Although nonlinear models perform well on training

data, a lack of proper regularization and cross-

validation can easily lead to overfitting, affecting the

accuracy of out-of-sample predictions. Therefore,

future research directions may include combining

deep learning with traditional financial theory to

improve the interpretability and applicability of the

model.

4 FUTURE OUTLOOK

Asset pricing methods are still evolving, and

emerging technologies and theoretical research are

driving the evolution of asset pricing models. Future

research directions mainly focus on behavioral

finance, data-driven methods, and improvements in

model limitations.

4.1 The Impact of Behavioral Finance

on Asset Pricing

Traditional asset pricing models assume that

investors are entirely rational, but studies have shown

that investor emotions and irrational behavior can

cause market prices to deviate from fundamental

values. Psychological biases such as herding,

overconfidence, and loss aversion can affect market

pricing and increase price volatility (Akin & Akin,

2024).

For example, during the pandemic, investors

generally followed market trends, and a large amount

of funds poured into the technology sector, pushing

stock prices far beyond their fundamental value.

Driven by market sentiment, investors ignored the

risk of overvaluation and formed a self-reinforcing

upward cycle. However, when market sentiment

reverses or actual earnings fall short of expectations,

the bubble bursts, and stock prices fall sharply,

ICDSE 2025 - The International Conference on Data Science and Engineering

256

reflecting the amplification of market volatility by the

herding effect (Yeh, Teoh, & Chu, 2020).

Behavioral finance corrects the rational

assumptions in traditional asset pricing theory and

provides a more reasonable explanation for market

price anomalies. Future asset pricing research can

incorporate behavioral finance into the model to

predict market trends more accurately.

4.2 Data-driven Asset Pricing Methods

With the advancement of big data technology, new

variables such as market sentiment and social media

sentiment analysis have become essential factors in

asset pricing. Traditional models mainly rely on

economic fundamentals data, while data-driven

methods can use unstructured data for market

forecasting.

Negative news sentiment can predict market

returns, and news analysis models combined with

natural language processing (NLP) technology can

improve the accuracy of asset pricing (Tetlock, 2007).

Market sentiment on social media platforms such as

Twitter can reflect investor sentiment fluctuations.

For example, the market sentiment index of social

media can predict stock market returns (Černevičienė

& Kabašinskas, 2024). The challenge of data-driven

methods lies in the problem of data falsification. For

example, information on social media may be false,

and data cleaning and feature selection are required to

improve the model's reliability.

4.3 Current Limitations

Although modern asset pricing models (such as

CAPM, APT, and multi-factor models) are widely

used, they still have limitations, such as the overly

idealistic assumption of rational investors, the

difficulty in capturing market nonlinear relationships,

the inability to explain market anomalies and the

neglect of market sentiment and non-financial factors.

At the same time, with the development of big data,

traditional models find it complex to use high-

dimensional data, and factor selection effectively

relies on experience, which affects the stability of

data processing.

Deep learning and machine learning provide new

opportunities for asset pricing, which can identify

complex nonlinear relationships, improve prediction

accuracy, and adapt to market dynamics. Machine

learning algorithms (SVM) perform well in market

return prediction and risk management, while deep

learning models automatically extract data features

through neural networks. However, these methods

still face challenges such as overfitting, black box

problems, and difficulty identifying causal

relationships. In the future, combining explainable

artificial intelligence, causal inference methods, and

multi-source data fusion is expected to improve

model transparency, optimize pricing accuracy, and

promote the transformation of financial analysis to

data-driven and intelligent (Černevičienė &

Kabašinskas, 2024).

5 CONCLUSIONS

Asset pricing is a complex and dynamic process

influenced by various external and internal factors.

Policy changes (such as monetary policy, fiscal

policy, and financial regulatory policy) directly affect

the formation of asset prices by changing market

liquidity, capital costs, and investor expectations.

Socioeconomic uncertainties (such as financial crises,

geopolitical conflicts, and natural disasters) further

affect the stability of asset pricing by exacerbating

market volatility and investor risk aversion. At the

same time, internal factors of enterprises (such as

financial status, management decisions, and

corporate governance structure) determine the

market's expectations of future corporate earnings,

which profoundly impacts asset valuation.

Traditional asset pricing models (such as CAPM,

APT, and multi-factor models) provide an important

theoretical basis for understanding the formation of

asset prices, but their limitations are also becoming

increasingly apparent. These models are usually

based on the assumption of rational investors, and it

is difficult to explain the impact of market anomalies

and irrational behaviors on asset prices. With the

rapid development of big data and artificial

intelligence technologies, applying machine learning

methods (such as neural networks and support vector

machines) in asset pricing has significantly improved

prediction accuracy. It can capture complex nonlinear

relationships and the characteristics of high-

dimensional data. However, machine learning models

still face challenges such as black box problems,

overfitting, and data quality, limiting their feasibility

in financial markets.

Future asset pricing research must combine

behavioral finance, data-driven methods, and

explainable artificial intelligence to reflect market

dynamics and investor behavior more

comprehensively. At the same time, research should

focus on improving the transparency and stability of

models in high-dimensional data environments,

avoiding overfitting, and enhancing the ability to

Multi-Dimensional Analysis and Exploration of Asset Pricing

257

explain market anomalies. In addition, with the

increase in global economic uncertainty, asset pricing

models need to integrate considerations of policy

changes and socioeconomic risks to improve their

applicability in actual investment decisions.

In short, asset pricing research is moving towards

multidisciplinary cross-integration. Technological

innovation and theoretical breakthroughs are

expected to provide investors with more accurate and

reliable analysis tools in the future, thereby

promoting the healthy development of financial

markets.

REFERENCES

Affes, W., & Jarboui, A. (2023). The impact of corporate

governance on financial performance: a cross-sector

study. Springer, 20(1), 374–394.

https://doi.org/10.1057/s41310-023-00182-8

Agnello, L., & Sousa, R. M. (2011). Fiscal policy and asset

prices. Bulletin of Economic Research, 65(2), 154–177.

Akin, I., & Akin, M. (2024). Behavioral finance impacts on

US stock market volatility: an analysis of market

anomalies. Behavioural Public Policy, 1–25.

Berkman, H., & Malloch, H. (2021). Stock valuation during

the COVID-19 pandemic: An explanation using option-

based discount rates. Journal of Banking & Finance,

106386.

Bernanke, B. S., & Kuttner, K. N. (2005). What explains

the stock market’s reaction to Federal Reserve policy?

The Journal of Finance, 60(3), 1221–1257.

Cai, K., & Rodavia, M. R. (2022, October). K-Means

Cluster Analysis Based on Consumer Behavior. In 2022

4th International Conference on Artificial Intelligence

and Advanced Manufacturing (AIAM) (pp. 143-146).

IEEE.

Chen, L., Pelger, M., & Zhu, J. (2024). Deep learning in

asset pricing. Management Science, 70(2), 714-750.

D'Souza, D. (2020). How war affects the modern stock

market. Investopedia. Retrieved from

https://www.investopedia.com/solving-the-war-

puzzle-4780889

Fang, M., & Taylor, S. (2021). A machine learning based

asset pricing factor model comparison on anomaly

portfolios. Economics Letters, 204, 109919.

Gallimberti, C. M., Lambert, R., & Xiao, J. J. (2021).

Banking market development and spillovers to stock

market liquidity. SSRN Electronic Journal.

https://doi.org/10.2139/ssrn.3788549

Gratton, P. (2024). Basel III: Get up to speed on the banking

rules behind all the commercials. Investopedia.

Retrieved from

https://www.investopedia.com/terms/b/basel-iii.asp

Hayes, A. (2020, November 17). Understanding arbitrage

pricing theory. Investopedia. Retrieved from

https://www.investopedia.com/terms/a/apt.asp

Hayes, A. (2024a, January 29). Fama and French three-

factor model definition: Formula and interpretation.

Investopedia. Retrieved from

https://www.investopedia.com/terms/f/famaandfrencht

hreefactormodel.asp

Jurgita Černevičienė, & Audrius Kabašinskas. (2024).

Explainable artificial intelligence (XAI) in finance: a

systematic literature review. Artificial Intelligence

Review, 57(8).

Kenton, W. (2024, July 1). Capital asset pricing model

(CAPM) and assumptions explained. Investopedia.

Retrieved from

https://www.investopedia.com/terms/c/capm.asp

Li, S. (2023). The impact of the Federal Reserve raising

interest rates on the US stock market index: The

dynamic perspective. Advances in Economics

Management and Political Sciences, 61(1), 265–274.

Li Xingyou. (2018). Comparative analysis of neural

networks and multi-factor models in quantitative

investment. Market Forum, 8.

Muir, T. (2017). Financial crises and risk premia. The

Quarterly Journal of Economics, 132(2), 765–809.

https://doi.org/10.1093/qje/qjw045

Pan Shuiyang. (2019). Big data, machine learning and asset

pricing. Modern Management Science, (2), 6-8.

Tetlock, P. (2007). Giving content to investor sentiment:

The role of media in the stock market. The Journal of

Finance, 62(3), 1139–1168.

Yeh, C.-C., Teoh, H.-J., & Chu, H.-H. (2020). Stock bubble

prediction and investor behavior bias: Application of

rough set theory. Lientung University International

Business Department. Retrieved from

https://finance.cyut.edu.tw/var/file/25/1025/img/834/2

020v2-3.pdf?utm_source=chatgpt.com

ICDSE 2025 - The International Conference on Data Science and Engineering

258