User-click Modelling for Predicting Purchase Intent (2112.02006v1)

Abstract: This thesis contributes a structured inquiry into the open actuarial mathematics problem of modelling user behaviour using machine learning methods, in order to predict purchase intent of non-life insurance products. It is valuable for a company to understand user interactions with their website as it provides rich and individualized insight into consumer behaviour. Most of existing research in user behaviour modelling aims to explain or predict clicks on a search engine result page or to estimate click-through rate in sponsored search. These models are based on concepts about users' examination patterns of a web page and the web page's representation of items. Investigating the problem of modelling user behaviour to predict purchase intent on a business website, we observe that a user's intention yields high dependency on how the user navigates the website in terms of how many different web pages the user visited, what kind of web pages the user interacted with, and how much time the user spent on each web page. Inspired by these findings, we propose two different ways of representing features of a user session leading to two models for user click-based purchase prediction: one based on a Feed Forward Neural Network, and another based on a Recurrent Neural Network. We examine the discriminativeness of user-clicks for predicting purchase intent by comparing the above two models with a model using demographic features of the user. Our experimental results show that our click-based models significantly outperform the demographic model, in terms of standard classification evaluation metrics, and that a model based on a sequential representation of user clicks yields slightly greater performance than a model based on feature engineering of clicks.

An Exploration of User-Click Modelling for Predicting Purchase Intent

This master's thesis investigates the application of user-click modeling to predict purchase intent, specifically for non-life insurance products. In the context of an increasingly digitized commercial landscape, understanding user behavior through machine learning offers valuable insights that can enhance customer engagement and conversion rates. The paper, authored by Simone Borg Bruun, employs two types of neural network models to explore the potential of user-click data.

Methodology and Models

The paper employs two principal feature-based models to capture and predict user behavior. The first model involves engineering non-temporal features from user interactions on a service-based business website, with a Feed Forward Neural Network (FFNN) applied to map these features to purchase intent. The second approach leverages a Recurrent Neural Network (RNN) with Long Short Term Memory (LSTM) cells, which utilizes a sequence of features representing user interactions over time. These models are contrasted against baseline models and a demographic-feature-based model, offering a comprehensive evaluation of their predictive capabilities.

Experimental Results

The paper presents significant findings that underscore the predictive power of historical user interaction data. By applying standard classification metrics to unseen test data, both the FFNN and RNN models demonstrated improved accuracy over baseline models. Intriguingly, the RNN which preserves the temporal order of interactions offers slightly superior performance by avoiding the need for manual feature engineering. This suggests that user interaction sequences carry critical temporal information that enhances prediction accuracy.

A comparative analysis reveals that click data exhibits greater discriminative capacity than demographic data in predicting purchase intent. This finding is pivotal as it highlights the inherent behavioral indicators within click data that remain robust across various devices and user contexts.

Implications and Future Research Directions

The implications of these findings are multifaceted. Practically, they advocate for integrating behavioral data into machine learning models used by businesses seeking to optimize online sales strategies. Theoretically, the research advances understanding of sequence-based modeling, encouraging deeper investigation into temporal dependencies in user data. Furthermore, the paper recommends exploring the integration of social network data and the development of persuasive recommendations to tailor sales strategies further.

Challenges related to privacy and ethics are also acknowledged, emphasizing the necessity for compliant data handling and user consent. The paper touches upon advanced privacy-preserving techniques such as differential privacy and k-anonymity, which remain critical considerations amidst evolving data protection regulations.

Future research could delve into social networking influences on purchase behavior and develop customized, persuasive digital experiences that align with individual customer journeys. These areas hold promise for refining user-click models and advancing personalized marketing methodologies.

In conclusion, this thesis offers a comprehensive exploration of user-click modeling, highlighting its efficacy and potential integration with broader demographic and social data to refine predictive analytics. It presents a valuable contribution to the field of machine learning-driven consumer behavior analysis, inviting further research into combining diverse data sources for enhanced predictive precision in e-commerce environments.