Machine-Learning-Based Approaches for Learning Marketing Strategies

Restricted (Penn State Only)
Lin, Yu San
Graduate Program:
Computer Science and Engineering
Doctor of Philosophy
Document Type:
Date of Defense:
October 05, 2018
Committee Members:
  • Wang-Chien Lee, Dissertation Advisor
  • Kamesh Madduri, Committee Member
  • Sencun Zhu, Committee Member
  • Wang-Chien Lee, Committee Member
  • Mary Beth Rosson, Outside Member
  • Wang-Chien Lee, Committee Chair
  • machine learning
  • deep learning
  • probabilistic generative model
  • marketing
  • crowdfunding
  • menu design
  • competition
  • word worth
  • reward
  • kickstarter
Studying the markets for better business strategies has been a pressing and practical issue. However, there is not enough attention paid into such research field when it comes to precise computational models. In this thesis, I looked into various marketing problems via approaches including data analysis, probabilistic generative models, and neural network models. The specific research problems I looked into includes (1) analyzing the rewards on crowdfunding platforms and how they correlate with crowdfunding project success, (2) learning the offering and bundling preferences of menu bundles in crowdfunding projects, (3) modeling the market competition and capturing the businesses' competitiveness, and (4) learning the word embedding that carries both the underlying semantic meaning and economic worths. We approach each of these research problems as follows. In the first paper, Analysis of Rewards on Reward-Based Crowdfunding Platforms, we analyze a Kickstarter dataset, which consists of approximately 3K projects and 30K rewards. The analysis employs various statistical methods, including Pearson correlation tests, Kolmogorov-Smirnow test and Kaplan-Meier estimation, to study the relationships between various reward characteristics and project success. We find that projects with more rewards, with limited and late-added rewards are more likely to succeed. We also categorize and automatically annotate rewards into fifteen reward item types. We further analyze how different types of reward items are adopted across various categories of projects as well as how they potentially affect project success. We discover that projects with include the previous rewards are most likely to succeed. Also, different categories of projects may employ different best strategies of adopting reward item types to achieve pledging goals. Finally, we verify the efficacy of reward-related information through predicting project success. The result shows that features extracted from rewards help better predict the successes of crowdfunding projects. In the second paper, Modeling Menu Bundle Designs of Crowdfunding Projects, we raise a novel research question: understanding project creators' decisions of reward designs to level their chance to succeed. We approach this by modeling the design behavior of project creators, and identifying the behaviors that lead to project success. We propose a probabilistic generative model, Menu-Offering-Bundle (MOB) model, to capture the offering and bundling decisions of project creators based on collected data of 14K crowdfunding projects and their 149K reward bundles across a half-year period. Our proposed model is shown to capture the offering and bundling topics, outperform the baselines in predicting reward designs. We also find that the learned offering and bundling topics carry distinguishable meanings and provide insights of key factors on project success. In our third paper, Modeling Dynamic Competition on Crowdfunding Markets, we study the competition on crowdfunding markets through data analysis, and propose a probabilistic generative model, Dynamic Market Competition (DMC) model, to capture the competitiveness of projects in crowdfunding. Through an empirical evaluation using the pledging history of past crowdfunding projects, our approach has shown to capture the competitiveness of projects very well, and significantly outperforms several baseline approaches in predicting the daily collected funds of crowdfunding projects, reducing errors by 31.73% to 45.14%. In addition, our analyses on the correlations between project competitiveness, project design factors, and project success indicate that highly competitive projects, while being winners under various setting of project design factors, are particularly impressive with high pledging goals and high price rewards, comparing to medium and low competitive projects. Finally, the competitiveness of projects learned by DMC is shown to be very useful in applications of predicting final success and days taken to hit pledging goal, reaching 85% accuracy and error of less than 7 days, respectively, with limited information at early pledging stage. In our fourth paper, Learning the Economic Worth of Words with Joint-Task Neural Network Model, we fill in this gap by proposing neural network models that learn a representation of words, which capture the underlying economic worths of words. The models also incorporate factors, e.g., the product's brand, and the restaurant's city, that may affect the aggregated economic values of an item through its textual description. We design an attention mechanism to further capture the different contribution of words at different positions in the sentences/phrases toward the aggregated economic value. Our models are shown to well predict not only the economic values when given the textual descriptions, but also the missing word when given other context words and the aggregated economic value. We design recommendation task to show that our models may serve as a great tool for word-choice recommendation system when a specific perceived economic value is targeted. In particular, in restaurant menu creation, our models recommend words to describe menu dishes that fit the desired perceived values.