The Data Science of Internet Economics: Modeling, Analysis and Inference

With the advancement of Internet technology, a variety of novel Internet services have emerged over the past decade, e.g., recommendation systems like IMDB, crowdsourcing systems like Amazon Mechanical Turck, E-commerce systems like eBay, etc. This thesis aims to explore the efficiency and effectiveness of such Internet services.;First, we study the efficiency and effectiveness of group recommendation systems. We consider a product recommendation system which recommends products to a group of users. The system only has partial preference information: a user only shows preference to a small subset of products in the form of rating. This partial preference information makes it a challenge to produce an accurate recommendation. We explore a number of fundamental questions. What is the desired number of ratings per product so as to guarantee an accurate recommendation? What are some effective voting rules in summarizing ratings? How users' misbehavior such as cheating may affect the recommendation accuracy? We propose a mathematical model to capture various important factors of a group recommendation system. We derive analytical expressions for recommendation accuracy measures and show that it is computationally expensive to evaluate them.;Second, we study the efficiency and effectiveness of online quality evaluation systems. Such systems interpret ratings as product quality assessments instead of personal preferences, e.g., Epinions. They provide historical product ratings so that users can evaluate the quality of products. Product ratings are important since they affect how well a product will be adopted by the market. The challenge is that we only have "partial information" on these ratings: each user assigns ratings to only a small subset of products. Under this partial information setting, we explore a number of fundamental questions: What is the "minimum number of ratings" a product needs so that one can make a reliable evaluation of its quality? How users' misbehavior such as cheating in product rating may affect the evaluation result? We propose a probabilistic model to capture various important factors (e.g., rating aggregation rules, rating behavior, etc.) that may influence the product quality assessment under the partial information setting. We derive the minimum number of ratings needed to produce a reliable indicator on the quality of a product. We extend our model to accommodate users' misbehavior in product rating. We derive the maximum fraction of misbehaving users that a rating aggregation rule can tolerate and the minimum number of ratings needed to compensate.;Third, we study the efficiency and effectiveness of eBay-like reputation systems, where ratings are interpreted as credits. E-commerce systems such as eBay and Taobao of Alibaba are becoming increasingly popular. Having an effective reputation system is critical to this type of Internet service because it can assist buyers to evaluate the trustworthiness of sellers, and it can also improve the revenue for reputable sellers and E-commerce operators. We formulate a stochastic model to analyze an eBay-like reputation system and propose four measures to quantify its effectiveness: (1) new seller ramp up time, (2) new seller drop out probability, (3) long term profit gains for sellers, and (4) average per seller transaction gains for the E-commerce operator. Through our analysis, we identify key factors which influence these four measures. Using a real-life dataset from eBay, we also discover that eBay suffers from long ramp up time, low long term profit gains and low average per seller transaction gains. We propose a new insurance mechanism which consists of an insurance protocol and a transaction mechanism to improve the above four measures.;Last, we design a class of incentive and reputation mechanisms for crowdsourcing systems to improve its efficiency. online crowdsourcing services are quite common such as Amazon Mechanical Turk, Elance and Yahoo! Answers. For such online services, it is important to attract "workers" to provide high-quality solutions to the "tasks" outsourced by "requesters". In this paper, we design a class of incentive and reputation mechanisms to solicit high-quality solutions from workers. Our incentive mechanism allows multiple workers to solve a task, splits the reward among workers based on requester evaluations of the solution quality, and guarantees that high-skilled workers provide high-quality solutions. Our reputation mechanism ensures that low-skilled workers do not provide low-quality solutions by tracking workers' historical contributions, and penalizing those workers having poor reputation. We show that coupling our reputation mechanism with our incentive mechanism, at least one high-quality solutions can be collected. (Abstract shortened by ProQuest.).