This study investigates the ex-ante information on initial public offering(IPO)underpricing in Malaysia using multiple machine learning techniques.The paper analyzes a sample of 350 fixed-price IPOs from 2004 to 2021,applying five machine learning models:artificial neural networks,random forest,gradient boosting,extra trees,and linear regression.The results indicate that random forests demonstrates superior performance,with a test R2 of 0.2292,a CV R2 mean of 0.529,and a CV R2 standard deviation of 0.1293,indicating moderate but reliable predictive power suitable for noisy financial data,such as IPO underpricing,where feature importance insights are more valuable than precise predictions.To reconcile and aggregate different outcomes from these multiple models,we implemented a voting algorithm to identify robust and reliable feature ranking for ex-ante determinants of IPO underpricing.Among the features,investor demand and divergence of opinions consistently emerged as the top two influential predictors of IPO underpricing,highlighting the key role investor sentiment and information asymmetry play in determining IPO pricing.These findings offer insights to investors,issuers,and policymakers,enabling a deeper understanding and effective management of the ex-ante drivers of underpricing in fixed-price IPOs,which lack market-based price discovery.
The research purpose is to contribute to the field of forecasting foreign exchange.This is due to the ever-changing economic conditions under which analysts can observe the significant volatility of exchange rate forecasts,as exchange rate forecasting has been challenging for analysts for many years.Stakeholders(central banks,governments,and investors)will seek to maximize asset returns and minimize risk in their decision-making using exchange rate forecasting.Therefore,this study proposes a new approach from the Black-Scholes model to forecast daily,weekly,monthly,and yearly exchange rates for the domestic currency pair the indonesian rupiah(IDR)to the United States dollar(USD)traded.The Black-Scholes model,originally referred to as a partial differential equation,was changed to an ordinary differential equation,which is used to approximate the numerical solution of the new Black-Scholes equation.The numerical solution used in this research is the fourth-order Runge-Kutta method.This study uses actual exchange rate data for more than one year,and the prediction results show that the proposed methodology can be an effective method for forecasting the exchange rate(IDR/USD).It is indicated by a small error,of less than 5％,and a mean absolute percentage error of less than 2％.
With the rapid advancement of the Internet of Things,the generation and sharing of massive data have become a significant trend.However,the pervasive free-riding behavior among stakeholders has adversely impacted data quality.To address this issue,this study employs tripartite evolutionary game theory to construct a decision-making model involving data providers,data brokers,and regu-lators.This model depicts the strategic choices of these three parties in data quality management:data providers and brokers may opt for proactive investment or passive free-riding,while regulators may choose between stringent oversight or routine inspections.By constructing replicator dynamics equations,employing Jacobian matrix analysis to examine equilibrium points,and combining numerical simulations with sensitivity analysis,this paper explores the evolution of stakeholder strategies and the impact of key parameters on system stability.Results indicate that free-riding behavior significantly compromises data quality.When the net benefit of active quality control falls below the free-riding payoff,the system converges toward a fully passive equilibrium point(0,0).However,appropriately calibrated incentive and penalty mechanisms can effectively promote active behavior.This study pro-vides a quantitative foundation for understanding multi-agent interactions in data sharing and offers actionable strategic insights for optimizing IoT data sharing mechanisms.
The development of the new energy vehicle(NEV)industry is pivotal in addressing China's energy security challenges and restructuring its automotive sector.It is a critical measure for achieving China's carbon neutrality goals.Accurately analyzing and forecasting NEV sales volumes carries sig-nificant implications for industry planning and policy formulation.This paper proposes an integrated modeling framework that combines variable selection techniques with the XGBoost algorithm to forecast NEV sales in China.We compare three distinct variable selection methods and evaluate their perfor-mance against commonly used benchmark forecasting models.The prediction performance is assessed using two metrics:root mean squared error(RMSE)and mean absolute percentage error(MAPE).Our empirical results demonstrate that the variable selection-XGBoost integrated model outperforms both univariate models and models that do not incorporate core factor extraction,showing superior accu-racy in both in-sample and out-of-sample predictions.Among the variable selection-XGBoost models,the SSL-XGBoost model yields the best performance,followed by GLMNET-XGBoost and LARS-XGBoost.The SSL method selects the greatest number of core variables.Cross-validation results indicate that integrating XGBoost with variable selection significantly reduces prediction errors.The variable selection-XGBoost integrated model surpasses traditional models in terms of both accuracy and reliability.
This study develops a systematic framework to evaluate regional service science and tech-nology innovation(SSTI)in China.It aims to address the"manufacturing paradigm"bias in existing evaluations.Drawing on the national innovation system(NIS)theory,a multi-level indicator system is constructed with three dimensions:Innovation input,innovation output,and innovation environment.The analytic hierarchy process(AHP)is applied to determine indicator weights,and the SSTI perfor-mance of eight provinces and municipalities in China from 2011 to 2023 is systematically measured and comparatively analyzed.The results reveal a sustained upward trend in the overall level of SSTI.However,a pronounced stratified pattern of"the strong get stronger"persists,reflecting significant and widening regional disparities.Moreover,regional innovation modes exhibit marked structural hetero-geneity.These findings advance the operational evaluation of SSTI and provide diagnostic insights for identifying systemic bottlenecks,optimizing resource allocation,and formulating region-specific innova-tion strategies.
