Informs Annual Meeting 2017

TA29

INFORMS Houston – 2017

TA29

TA30

350E Artificial Intelligence Problems in Real World Sponsored: Artificial Intelligence Sponsored Session Chair: Jiaheng Xie, University of Arizona, 1130 E Helen Street, Tucson, AZ, 85721, United States, xiej@email.arizona.edu 1 - Genetic Algorithms for Electrical Harness Routing During Satellite Conception Eliott Roynette, PhD Student, Airbus Defence and Space, In the space industry, every satellite is unique and customized for its mission. In this presentation, we introduce a multi-criteria genetic algorithm for helping in the conception of the 3D routing of the electrical harness, which is responsible for establishing physical connections between several thousands of devices of the satellite. This algorithm is used for obtaining a tradeoff between the total weight of the harness and the complexity of its manufacturing. In the end, we show how using OR tools brings a strong advantage to space engineers both for speeding and optimizing the conception process of real satellites. 2 - Forecasting Crude Oil Price: A Novel Decompose-ensemble Methodology with AIC-ANN Approach Yishan Ding, Shanghai Jiaotong University, Huashan Road 1954, Xuhui District, Shanghai Jiaotong University, Zhongyuan, Shanghai, 200030, China, dys90@163.com Forecasting international crude oil is a well-known issue. Hybrid methodology has been widely used in current research. In this study, a novel decompose- ensemble prediction process is proposed. This method, adds the decompose-ensemble to the single AI model to further improve the predicting accuracy. It has been confirmed that “decompose-ensemble” models are better than the normal hybrid one, in terms of prediction accuracy (both level and directional measurement) and Diebold-Mariano test. What’s more, the EEMD- based one outperforms the empirical mode decomposition(EMD) one. At last but not the least, AIC gives us reasonable and convincing statement about determining the value of lag. 3 - Advantages of the Parameter Control Strategy Over the Parameter Tuning: A Case Study for the Berth Scheduling Problem at Marine Container Terminals Maxim A.Dulebenets, Florida A&M.University-Florida State University, 2300 Bluff Oak Way, Apt. 8408, Tallahassee, FL, 32311, United States, mdlbnets@gmail.com, Eren Erman Ozguven, Ren Moses, Thobias Sando This study proposes a novel Evolutionary Algorithm to assist with berth scheduling at marine container terminals, which unlike published to date studies on berth scheduling applies a parameter control strategy. Specifically, an adaptive mechanism is developed for the mutation operator, where the mutation rate is altered based on feedback from the search. A set of numerical experiments are conducted to assess performance of the developed algorithm based on a comparison against a typical Evolutionary Algorithm that relies on the parameter tuning analysis. 4 - Path Planning for Multiple UAVs in Dynamic Environment Considering Uncertainty of Threats Mengqi Hu, University of Illinois at Chicago, 842 W. Taylor Street, MC 251, 3023 ERF, Chicago, IL, 60607, United States, mhu@uic.edu, Di Wang Path planning for multiple interconnected unmanned aerial vehicles (UAVs) especially for fixed-wing UAVs in dynamic environment has attracted great attention. Considering motion characteristics of dynamic artificial and natural threats, a probability predictive method based on Bayesian theory is presented to modify cost functions. A non-cooperative game model is formulated in which UAVs are modeled as Homo Egualis agents that aim at predicting a sequence of future actions and then a novel reinforcement learning method is proposed to solve the game model to converge to a Nash equilibrium. Finally, the feasibility and effectiveness of the proposed approach is evaluated using simulations. 31 Rue des Cosmonautes, Toulouse, 31400, France, eliott.roynette@gmail.com, Cedric Pralet, Vincent Vidal, Bertrand Cabon

350F Topics in Social Media Analytics II Invited: Social Media Analytics Invited Session Chair: Fouad H. Mirzaei, Santa Clara University, Santa Clara, CA, 95050, United States, fhmirzaei@scu.edu 1 - Attracting Investors in Equity Crowdfunding Campaigns Anna Lukkarinen, Aalto University, Helsinki, 00100, Finland, anna.lukkarinen@aalto.fi, Jyrki Wallenius, Hannele Elina Wallenius, Tomi Seppälä While many equity crowdfunding campaigns succeed in reaching their funding goals, many also fail. Using campaign and investor level data, we aim to build an understanding of what attracts the crowd to invest. We find that campaign characteristics and networks are more important for success than traditional factors. Furthermore, investors form distinct clusters that differ in their motivations and decision criteria. Donors and supporters place importance on personal acquaintance, while pure investors look at the target company and campaign features. The findings can assist crowdfunding platforms and entrepreneurs in selecting and targeting relevant segments of potential investors. 2 - Social Media Competitiveness Fouad H. Mirzaei, Santa Clara University, Unit 3, 559 Alviso St, Santa Clara, CA, 95050, United States, fhmirzaei@scu.edu In this study, we investigate what factors are vital in the short term and the long term success of social media networks. 3 - Do Financial Incentives Induce More Online Participatory Behaviors? Lini Kuang, Beijing Institute of Technology, 5TH, Zhongguan Online community-based Question and Answer (Q&A) websites is popular in recent years. The main challenge for Q&A websites is how to motivate users’ participation. Using a natural experimental dataset collected from a Chinese Q&A site, we apply difference-in-differences model with propensity score matching to evaluate the impact of paid knowledge sharing activity on free knowledge sharing behaviors and social activities in Q&A sites. Our results show that the paid activity induces more free knowledge sharing and social ties in the site. The results suggest that the financial incentive can increase users’ participatory behaviors significantly. 351A Finance Contributed Session Chair: Yen-Ju Chiang, National Cheng Kung University, Tainan, Taiwan, r48031016@mail.ncku.edu.tw 1 - Corporate Bankruptcy Prediction a Penalized Semiparametric Index Hazard Model Approach Shaobo Li, PhD Student, University of Cincinnati, 5470 Erie Station Lane, Apt.51, Cincinnati, OH, 45227, United States, lis6@mail.uc.edu, Shaonan Tian, Yan Yu We introduce a parsimonious yet flexible double-index hazard model for corporate bankruptcy prediction under a semiparametric modeling framework. The two indices are naturally constructed by market and accounting based variables respectively. Penalized estimation is considered variable selection. Our empirical results demonstrate the existence of nonlinearity between predictors and default risk, and improvement of predictive power comparing to Shumway’s hazard model. Furthermore, we find that the variables in each index vary as the prediction horizon changes. The proposed model framework sheds new light on modeling flexibility and sparsity in various financial applications. 2 - Efficient Hybrid Candlestick Technical Analysis Model for Stock Market Timing on the Basis of the Support Vector Machine and Heuristic Algorithms Milad Jasemi, Wayne State University, 4815 4th Street, Detroit, MI, 48201, United States, milad@wayne.edu, Leslie Monplaisir, Elham Ahmadi Here, two hybrid models are used for timing of stock markets based on the technical analysis of Japanese Candlestick by Support Vector Machine (SVM) and Heuristic Algorithms of Imperialist Competition and Genetic. In the first model, ICA is used to optimize the SVM parameters’. In the second model, Genetic Algorithm is used for feature selection in addition to SVM parameters’ optimization. To generate the input data, the Raw-based and Signal-based standard approaches of the literature are applied. According to the Hit Rate for periods of 1-6 day, SVM-ICA prevails them all. Street, Haidian District, Beij, Beijing, 100081, China, 3120160689@bit.edu.cn, Zhijun Yan, Han Yang, Yan Liu TA31

280