MM will hold a Joint Public Lecture on the topic “Unveiling E-Book Bestsellers: Predictive Insights from Consumption Patterns Using Machine Learning” and “Introducing Machine-Learning-Based Data Fusion Methods for Analyzing Multimodal Data: An Application of Measuring Trustworthiness of Microenterprises” on 19 June 2024.
Unveiling E-Book Bestsellers: Predictive Insights from Consumption Patterns Using Machine Learning
Abstract:
In the e-book industry, the pro tability of stakeholders hinges upon the sales of “bestsellers.” However, accurately identifying these titles poses a signi cant challenge. While early online reviews have traditionally served as a key resource for forecasting new e-book sales, their reliability and credibility are often undermined by concerns, such as manipulation, rating in ation, and “cold-start” issues. As a result, sales predictions solely based on consumer-generated reviews may fall short of expectations. Our study introduces innovative consumption-based prediction approaches that leverage readers’ consumption patterns, o ering a promising avenue for identifying bestsellers. Drawing from the perspectives of sustained attention, we identify three key aspects of users’ consumption trajectories— amount, duration, and intensity— and incorporate them into our prediction models established using time-sequenced machine learning algorithms, such as LSTM. Our ndings demonstrate substantial improvements in bestseller prediction accuracy when consumption data is integrated into the models alongside online review and book characteristic parameters. Consumption-driven predictions signi cantly enhance sensitivity compared to baseline models. The highest performance enhancement is achieved when both consumption and online reviews are considered in tandem. Furthermore, we explore the nuanced impact of consumption-based predictions across various review characteristics, such as valence, extremity, and informativeness. In addition, the inclusion of consumption data can be particularly helpful for predicting the success of e-books by new authors with no historical sales. Through rigorous robustness checks, we validate the reliability of our ndings, a rming that consumption-driven and online review-based predictions e ectively complement and substitute each other in enhancing sales forecasts for digital content products.
Introducing Machine-Learning-Based Data Fusion Methods for Analyzing Multimodal Data: An Application of Measuring Trustworthiness of Microenterprises
Abstract:
Multimodal data, comprising interdependent unstructured text, image, and audio data that collectively characterize the same source, with video being a prominent example, o er a wealth of information for strategy researchers. Our study highlights the vital role of both verbal and nonverbal communication in attaining strategic objectives. Through the analysis of multimodal data—incorporating text, images, and audio—we demonstrate the essential nature of interpersonal interactions in bolstering trustworthiness, thus facilitating the success of microenterprises. Leveraging advanced machine learning techniques, such as data fusion for multimodal data and Explainable Arti cial Intelligence (XAI), we notably enhance predictive accuracy and theoretical interpretability in assessing trustworthiness. By bridging strategic research with cutting-edge computational techniques, we provide practitioners with actionable strategies for enhancing communication e ectiveness and fostering trust-based relationships. Access our data and code for further exploration.
Date* : 19 June 2024 (Wednesday)
Time* : 2 to 3:50 pm
*Hong Kong Time
Venue : N001, PolyU campus
Please register through the registration form on or before 16 Jun 2024 (Sun).
