MACHINE LEARNING APPROACHES FOR DEMAND FORECASTING: THE IMPACT OF CUSTOMER SATISFACTION ON PREDICTION ACCURACY

Аннотация

INTRODUCTION In an era characterized by rapid technological Machine learning (ML) offers a robust framework advancements and shifting consumer preferences, for analyzing large and complex datasets, allowing the ability to predict product demand accurately businesses to leverage customer feedback, has become a critical factor for businesses striving reviews, and satisfaction scores to enhance to maintain a competitive edge. Effective demand demand predictions. Unlike traditional methods, forecasting not only helps organizations manage ML algorithms can adapt to new information, inventory efficiently but also enhances customer continuously learning from data to improve satisfaction by ensuring that products are available accuracy over time. This adaptability is when and where customers need them. Traditional particularly valuable in today’s fast-paced market forecasting methods often rely on historical sales environment, where consumer preferences can data and simplistic statistical models, which may change rapidly and unpredictably. fail to capture the complexities of consumer The aim of this study is to investigate the behavior and the myriad factors influencing effectiveness of various machine learning models demand. in predicting product demand based on customer Customer satisfaction has emerged as a pivotal satisfaction data. By exploring different algorithms determinant of demand, reflecting consumers' and assessing their performance using rigorous perceptions of product quality, service levels, and evaluation metrics, this research seeks to identify overall experience. Research indicates that the most effective approach for businesses seeking satisfied customers are more likely to become to optimize their demand forecasting processes. repeat buyers, leading to increased sales and Ultimately, the findings of this study aim to improved brand loyalty. As such, integrating contribute valuable insights to both academic customer satisfaction data into demand literature and practical applications, guiding forecasting models can provide a more nuanced organizations in making informed decisions that understanding of market dynamics. However, enhance operational efficiency and customer harnessing this data effectively requires satisfaction. sophisticated analytical techniques that can Literature Review uncover hidden patterns and relationships. The intricate relationship between customer https://www.theamericanjournals.com/index.php/tajet 43 THE USA JOURNALS THE AMERICAN JOURNAL OF ENGINEERING AND TECHNOLOGY (ISSN – 2689-0984) VOLUME 06 ISSUE10 satisfaction and product demand has been the these advanced modeling techniques can yield subject of extensive research across various significant improvements in demand forecasting disciplines, including marketing, operations, and accuracy. data analytics. Numerous studies have established Furthermore, the importance of model a positive correlation between customer interpretability has gained traction in the machine satisfaction and subsequent purchasing behavior, learning community, particularly in contexts reinforcing the idea that satisfied customers drive where decision-makers need to understand the higher demand. For instance, Anderson and Mittal factors driving model predictions. Lundberg and (2000) suggest that businesses that prioritize Lee (2017) introduced SHAP (SHapley Additive customer satisfaction can expect increased repeat exPlanations), a method that provides insights into purchases and stronger brand loyalty. These how individual features contribute to model findings highlight the importance of outputs. This transparency is crucial for understanding customer sentiments as a critical organizations looking to leverage machine component of effective demand forecasting. learning for demand forecasting, as it allows In recent years, the advent of machine learning practitioners to make informed decisions based on techniques has revolutionized demand forecasting the factors influencing predictions. by enabling businesses to leverage vast amounts of The current study seeks to build upon this data for predictive analytics. Traditional extensive literature by conducting a forecasting methods, such as exponential comprehensive evaluation of various machine smoothing and moving averages, often struggle to learning models for predicting product demand capture the complexities and non-linear based on customer satisfaction metrics. By relationships present in consumer behavior. employing a systematic approach to model Hyndman and Athanasopoulos (2018) argue that selection, training, and evaluation, this research these limitations necessitate the adoption of more aims to identify the most effective algorithms for advanced methodologies, including machine capturing the nuances of consumer behavior and learning algorithms, which can model intricate improving demand forecasts. Ultimately, the patterns and adapt to changes in consumer findings will provide valuable insights