Yongli Li

Exploring the Characteristics of Innovation Adoption in Social Networks: Structure, Homophily, and Strategy

Exploration of the characteristics of innovation adoption in the context of social network will add new insights beyond the traditional innovation models. In this paper, we establish a new agent-based model to simulate the behaviors of agents in terms of innovation adoption. Specifically, we examine the effects of the network structure, homophily and strategy, among which homophily is a new topic in this field of innovation adoption. The experiments illustrate six important findings involving five aspects and their influences on the innovation adoption. The five aspects are initial conditions, homophily, network topology, rules of updating and strategy, respectively. This paper also compares the different cases within one aspect or across several aspects listed above. Accordingly, some management advices and future work are provided in the last part of this paper.

A stochastic DEA model considering undesirable outputs with weak disposability

Abstract This paper proposes a stochastic DEA model considering undesirable outputs with weak disposability which not only can deal with the existence of random errors in the collected data, but also depicts the production rules uncovered by weak disposability of the undesirable outputs. This model introduces the concept of risk to define the efficiency of decision making units (DMUs), and utilizes the correlationship matrix of all the variables to portray the weak disposability. On the basis of probability distribution properties, the probabilistic form of the model is transformed to the equivalent deterministic one which is able to be solved. In the application of the model, the environment efficiency evaluation problem is chosen to validate the model by designing different levels of random errors and comparing the new model with the old one. In conclusion, the model, with broad applicability has a more superior analysis capacity than the existing model.

A network-based and multi-parameter model for finding influential authors

This study proposes a network-based model with two parameters to find influential authors based on the idea that the prestige of a whole network changes when a node is removed. We apply the Katz–Bonacich centrality to define network prestige, which agrees with the idea behind the PageRank algorithm. We further deduce a concise mathematical formula to calculate each authors influence score to find the influential ones. Furthermore, the functions of two parameters are revealed by the analysis of simulation and the test on the real-world data. Parameter α provides useful information exogenous to the established network, and parameter β measures the robustness of the result for cases in which the incompleteness of the network is considered. On the basis of the coauthor network of Paul Erdos, a comprehensive application of this new model is also provided.

A Combination Prediction Model of Stock Composite Index Based on Artificial Intelligent Methods and Multi-Agent Simulation

AbstractPredicting stock composite index is useful, which can raise the interest of both the investors and the corresponding researchers. This paper presented a new combination prediction model based on the technique of artificial intelligence and the principle of combination forecast. The principle of combination forecast, as a valid foundation of the new model, was strictly proved and carefully illustrated in this paper. Given the predicting rules, the new combination model was established by synthesizing three commonly used prediction models based on the principle of combination forecast. The comprehensive usage of qualitative forecast and quantitative forecast is also a feature of the new model. To valid the new model, comparison analysis and multi-agent simulation were both applied. Besides, the application of multi-agent simulation made the new model able to guide the investors’ operations in a real stock market. According to the theoretical proof, the comparison analysis and the simulation experime...

Rating online commodities by considering consumers’ purchasing networks

Purpose – Rating, as a common way of evaluation, is a significant exercise and plays a major role in managerial decision-making in general and in particular online purchasing. The paper aims to discuss these issues. Design/methodology/approach – This study utilizes the theory of social network analysis (SNA) to make a comprehensive evaluation model for rating commodities. Specifically, the paper shows how to apply the network analysis, how it works and what the advantage is. The paper further presents the new models properties and validates the models applicability. The paper finally analyzes the results with respect to various dimensions of a movie rating database and report on the insights generated by the model. Findings – Through the designed comparison analysis and the empirical analysis, the model is showed to be better than the traditional ones such as averaging, analytic hierarchy process (AHP) and several mentioned dimension-reduction techniques (DRTs) in terms of consistency and its ability to...

Evaluation and recommendation methods based on graph model

Evaluation and recommendation are different actions, but they are consistent in mining and using information efficiently and effectively to improve their persuasiveness and accuracy. From the view of information processing, the paper builds a two-dimensional graph model which expresses the relationships between evaluators and objects. This graph model reflects the original information of evaluation or recommendation systems and has its equivalent matrix form. Next, the principle of matrix projection can be applied to get the evaluation or recommendation vector by solving the matrix maximization problems.Whats more, a rating data set of online move is selected to verify the model and method. In conclusion, from the example analysis, it is found that the proposed evaluation method is reasonable, and from the numerical experimental comparison, the proposed recommendation method is proved to be timesaving and more accurate than the generally adopted recommendation methods.