BML : : Contributions

An Advanced Forecasting Model Leveraging Emotion‐Gesture Correlation to Predict Returning Visitors Surpasses Visit Duration as a Predictive Factor

This research paper explores the effectiveness of user emotional experience as a predictor for future returning of the user to the website. Traditional web analytics have been limited in their ability to accurately capture the nuances of user experiences. Methods like eye tracking, speech tracking, and surveys, while insightful, often suffer from being overly intrusive, leading to biased results. This study introduces a state of the art, nonintrusive method of measuring user experience: touch-gesture based emotion measurement. This technique leverages the subconscious nature of touch gestures to gather emotional data, allowing for a more authentic and unbiased user interaction with websites. To leverage this method, we first gained explicit data collection consent and gathered browsing data from 164,527 users across a 1-year period on a Canadian e-commerce website visited from a touchscreen device. While the sample size is significantly large, the sample is primarily made up of visitors from Canada, which could limit generalization of the findings. Using this data, we implemented an AI model which predicts whether a user is likely to return to the website or not, primarily based on their emotional touch gestures on their first visit with an accuracy of 91.7%. This approach not only enhances our understanding of user engagement but also opens new avenues for optimizing user experience in untested digital spaces such as e-learning and mental health.

Predicting user engagement levels through emotion-based gesture analysis of initial impressions

This study validates the predictive power of emotion-based gesture analysis in deter–

mining user engagement levels based on their initial impressions of a website. To

achieve the objective of this research, we conducted experiments for 53weeks by

capturing the data from 3,797 unique visitors engaging with an e-commerce web–

site. Gesture-based emotion analysis was employed to capture users’ initial impres–

sions, encompassing gestures like clicks, scrolls, swipes, and taps on mobile touch

surfaces. Emaww AI’s proprietary first impression metric captures the initial emo–

tional response to a product or service and this initial impression can have a sig–

nificant impact on their overall impression of the site. The result of this research

demonstrates a strong correlation between five levels of users’ first impression to

the subsequent engagement during their visit to a website. The validated regression

model demonstrated commendable performance, as indicated by a Mean Absolute

Error (MAE) score of 1.60. This relatively low MAE suggests that the model’s pre–

dictions closely align with the actual values, reflecting a high level of accuracy. The

implications for diverse users highlight the importance of aligning website content

and aesthetic design with users’ emotions to drive engagement.

Improving User Engagement in E-Learning: A Machine Learning Approach Using Gesture-Based Emotion Tracking to Promote Mental Well-being

E-learning websites have transformed access to education by providing flexible, globally accessible environments for studying. Their effectiveness is often measured by user receptivity, which refers to how well users engage with and absorb the content presented. Prior research has focused primarily on multimedia integration, content organization, and interactive features as contributors to a successful e-learning experience. However, limited studies have investigated how the design and structure of these websites, in conjunction with real-time user engagement, influence attention and overall user experience. This research explores the impact of design and structural elements on user receptivity in e-learning websites, integrating real-time user engagement data with AI models to provide actionable insights. Using the Emotions dataset of over 6,000 users across 35 countries collected over 53 weeks via the Emaww API, we developed machine learning models to predict user receptivity based on website parameters like color schemes, number of images and paragraphs, and time spent on each page. The Light Gradient Boosting Machine (LightGBM) achieved the highest accuracy at 83.20%, with time spent on the website emerging as a key factor in improving model performance. Our findings offer concrete recommendations for optimizing e-learning platforms by enhancing design, layout, and content presentation, ultimately leading to improved user engagement and educational outcomes.