Journal of Computer Technology and Applied Mathematics

Open Access Research ArticlePublished 4 November 2025

Daily Asset Pricing Based on Deep Learning: Integrating No-Arbitrage Constraints and Market Dynamics

Yinlei Chen

We propose a novel deep learning approach to asset pricing that predicts individual stock returns using daily data while integrating no-arbitrage constraints and capturing market dynamics. Our model combines Long Short-Term Memory (LSTM) networks, Convolutional Neural Networks (CNN), and Generative Adversarial Networks (GAN) to model the complex relationships between stock returns and various conditioning variables, including macroeconomic indicators, technical indicators, and market sentiment data. By incorporating the no-arbitrage condition into the deep learning framework, we enhance the accuracy and stability of asset pricing. We estimate a stochastic discount factor that explains asset returns from the conditional moment constraints implied by no-arbitrage. Our method outperforms traditional multi-factor models, such as the Fama-French model, in terms of Sharpe ratio, explained variation, and pricing errors. The GAN enforces the no-arbitrage constraint by identifying portfolio strategies that contain the most pricing information. The LSTM network uncovers hidden economic states, while the feedforward network captures the non-linear effects of conditioning variables. This research provides a new direction in asset pricing by applying deep learning to integrate market dynamics and enforce no-arbitrage constraints, offering more accurate pricing and valuable insights for generating profitable investment strategies.

Disciplines:Artificial Intelligence and IntelligenceSubjects:Deep LearningReferences:20

Open Access Research ArticlePublished 4 November 2025

Application Development with AI: a Case Study of Application x Development Progress

Artificial intelligence (AI) has taken a leading position in various fields, especially in the development of information solutions. This article presents a case study of mobile application user interface redesign to highlight the benefits and improvements of integrating artificial intelligence. The study analyzes a typical mobile application, including technical and business details. The integration of artificial intelligence tools such as open source SolidGPT helps improve the development process in every aspect compared to the original version. The results show a lot of time saved and a lot of time reduced for similar activities. Additionally, the quality of the final product has improved, as evidenced by the reduction in crashes and issues found in similar time frames. The improvements achieved are very promising. The case study can be useful for technicians implementing AI tools and for decision-makers justifying investments in such integration.

Disciplines:Software TechnologySubjects:Software EngineeringReferences:9

Open Access Research ArticlePublished 4 November 2025

Prophet with Exogenous Variables for Procurement Demand Prediction under Market Volatility

Addressing the issue of insufficient accuracy in procurement demand forecasting under market volatility, this study investigates the Prophet model with exogenous variables. It outlines the comprehensive workflow encompassing data preprocessing, feature reconstruction, and model training, while introducing trend decomposition and forecasting implementation methods constrained by multi-source features. Comparative experimental results demonstrate that the improved model reduces RMSE by 21.5% and MAPE by 34.2% in high-volatility intervals, significantly enhancing prediction stability. This validates the effective corrective role of exogenous variables in addressing complex market disturbances.

Disciplines:Big Data TechnologySubjects:Data AnalyticsReferences:5

Open Access Research ArticlePublished 4 November 2025

LSTM-Based Deep Learning Models for Long-Term Inventory Forecasting in Retail Operations

Given the complex fluctuations and extended forecasting cycles inherent in retail inventory, this study investigates the application of LSTM deep learning models for inventory time series forecasting. It details the model architecture design, parameter optimization, and training methodology, while presenting the data construction and experimental validation process. Results demonstrate that the model effectively captures inventory variation patterns, enhances prediction accuracy and trend stability, and exhibits strong generalization capabilities across multiple forecasting horizons.

Disciplines:Artificial IntelligenceSubjects:Deep LearningReferences:5

Open Access Research ArticlePublished 4 November 2025

Funding-Rate Driven Routing in Financial Prediction Systems

Xueyi Cheng

Financial forecasting models tend to be based on past market action trends. While such methods are very good under steady-state conditions, financial markets tend to change abruptly due to a change in liquidity, macroeconomic news, or leverage imbalances. Such models trained on past price action are too sluggish in responding to regime changes and therefore make the wrong and not to be relied upon predictions at the worst possible moment when accurate advice is most required. To address this challenge, this paper proposes a funding-rate driven routing model that dynamically switches between forecasting models depending on inferred market states. Funding rates and spot–perpetual basis signs are used to detect volatility and sentiment changes that cause shock regimes. In ordinary times, a light univariate LSTM is used to ensure forecasting efficacy. As volatility rises, the system pushes forecasting to a multi-source deep learning model that incorporates price, volume, and sentiment background. The architecture uses an edge–cloud hybrid execution strategy that compromises on latency, privacy, and scalability. This work demonstrates that, through the integration of regime detection and model switching, one can obtain more accurate forecasting without sacrificing computational responsibility. The model introduces an additional more adaptive, context-aware methodology for real-world financial forecasting environments that must operate under continuously varying risk levels.

Disciplines:Big Data TechnologySubjects:Data AnalyticsReferences:10

Open Access Research ArticlePublished 4 November 2025

Fostering Deep Belonging Through Culturally-Responsive AI Mentorship Agents: An Identity-Affirming Framework for Educational Support

,

Culturally-responsive AI mentorship agents represent a substantial change in educational technology, addressing critical gaps between personalized learning systems and students' psychological needs for belonging. This research presents a comprehensive framework integrating multi-agent architectures with identity-affirming interaction mechanisms to cultivate deep belonging across diverse student populations. Through a mixed-methods empirical study involving 120 participants from three distinct cultural groups over six months, we demonstrate that culturally-adaptive AI mentors achieve 34.7% higher belonging scores compared to culturally-neutral systems. The framework employs dynamic cultural profiling, identity-safe feedback strategies, and personalized belonging interventions adapted from established psychological research. Statistical analysis reveals significant mediation effects where cultural responsiveness influences academic outcomes through enhanced belonging (β = 0.412, p < 0.001). Implementation across educational contexts shows differential effectiveness patterns, with underrepresented groups experiencing 42.3% greater benefit from culturally-responsive features. This work establishes theoretical foundations and practical guidelines for deploying AI mentorship systems that strengthen rather than diminish human connection in digital learning environments.

Disciplines:Artificial Intelligence and IntelligenceSubjects:Machine LearningReferences:45

Open Access Research ArticlePublished 4 November 2025

Sales Resource Optimization Based on Customer Lifetime Value (CLV): A Data-Driven Dynamic Decision Model

This study proposes a framework for a dynamic decision-making model based on Customer Lifetime Value (CLV) to optimize sales resources. This model prioritizes profitability as a benchmark for customer performance, serving as a guide for effective resource allocation and improvement. The paper presents a data-driven approach that combines real-time data and dynamic analysis of customer behavior for resource allocation and adjustment. This method analyzes enterprise data and customer behavior to inform decisions that maximize resource utilization and enhance sales performance. The core idea is to prioritize high-value customers and allocate more resources accordingly. Furthermore, by categorizing customers into different CLV groups, more informed and targeted decisions are made for each customer group. Ultimately, this improves customer retention, thereby enhancing the enterprise's long-term attractiveness and strategic model.

Disciplines:Statistical AnalysisSubjects:Data MiningReferences:14