Metaheuristics in Big Data: An Approach to Railway Engineering Silvia Galván Núñez 1,2, and Prof. Nii Attoh-Okine 1,3 1 Department of Civil and Environmental Engineering University of Delaware, Newark, DE, USA 2 silgalnu@udel.edu, 3 okine@udel.edu IEEE Workshop on Large Data Analytics in Transportation Engineering October 27, 2014 October 27, 2014 1 / 17
Outline 1 Introduction 2 Big Data and Optimization 3 Big Data in Railway Engineering 4 Conclusion and Future Work October 27, 2014 2 / 17
Section 1 Introduction October 27, 2014 3 / 17
Introduction Big Data and Optimization Big Data in Railway Engineering Conclusion and Future Work Introduction Huge amount of data generated on the internet every minute: Facebook users share over 2.5 million pieces of content. Email users send over 200 million messages. Google receives 4 million queries. http: //www.domo.com/blog/2014/04/data-never-sleeps-2-0/ October 27, 2014 4 / 17
Introduction Railway Engineering 3 terabytes of data are generated in a year for bearing temperature detectors for Class I US railroad [Li et al., 2014]. Data collected from different sensors need to be integrated for predicting track failures [Xie and Liu, 2010]. October 27, 2014 5 / 17
Introduction Figure 1: Five V s of Big Data ([Costa, 2013], [Lusher et al., 2013], [Cheng et al., 2013], [Pandey and Nepal, 2013], [Markowetz et al., 2014], [Chen, 2014]) October 27, 2014 6 / 17
Big Data Analytics Techniques - Some Examples Figure 2: Example of a framework for Big Data Analytics ([Li et al., 2013], IEEE [Tannahill Workshop and onjamshidi, Large Data2014]) Analytics in Transportation Engineering, October 27, 2014 7 / 17
Big Data Analytics Techniques - Some Examples Figure 2: Example of a framework for Big Data Analytics ([Li et al., 2013], IEEE [Tannahill Workshop and onjamshidi, Large Data2014]) Analytics in Transportation Engineering, October 27, 2014 7 / 17
Big Data Analytics Techniques - Some Examples Figure 2: Example of a framework for Big Data Analytics ([Li et al., 2013], IEEE [Tannahill Workshop and onjamshidi, Large Data2014]) Analytics in Transportation Engineering, October 27, 2014 7 / 17
Big Data Analytics Techniques - Some Examples Figure 2: Example of a framework for Big Data Analytics ([Li et al., 2013], IEEE [Tannahill Workshop and onjamshidi, Large Data2014]) Analytics in Transportation Engineering, October 27, 2014 7 / 17
Big Data Analytics Techniques - Some Examples Figure 2: Example of a framework for Big Data Analytics ([Li et al., 2013], IEEE [Tannahill Workshop and onjamshidi, Large Data2014]) Analytics in Transportation Engineering, October 27, 2014 7 / 17
Section 2 Big Data and Optimization October 27, 2014 8 / 17
Optimization Techniques Figure 3: Classification of Optimization Techniques October 27, 2014 9 / 17
Optimization Techniques Figure 3: Classification of Optimization Techniques October 27, 2014 9 / 17
Big Data Analytics using Optimization Techniques Optimization Technique Evolutionary optimization Ant colony optimization Authors [Fan et al.,2000],[chen, 2008], [Lee et al., 2012], [Chen et al., 2013], [Cambria et al., 2013], [Thomas and Jin, 2013], [Zhu et al., 2014], [Balicki et al. 2014], [Lee et al., 2014], [Liu, 2014],[Tannahill and Jamshidi, 2014] [Yang and Chen, 2006], [Sun et al., 2013], [Wu et al., 2013], [Zhang et al.,2014] Optimization Technique Particle swarm optimization Greedy algorithm Authors [Chaari et al., 2012], [Chang et al., 2013], [Fong et al., 2013], [Govindarajan et al., 2013], [Cheng et al., 2013], [Jian and Wang, 2014] [Chung et al., 2013], [Mestre and Pires, 2013], [Tan et al., 2013], [Lin et al., 2013], [Wang et al., 2014] Hierarchical Neighbor [Liu et al., 2013] [Yang and Fong, 2013] scheduling embedding l1-regularized [Saha et al., 2013], Horizontal data [Bellatreche et al., 2013] optimization [Tran et al., 2013] partitioning Method of multipliers 2014] annealing [Liu et al., 2013], [Anbari et al., Simulated [Rahimian et al., 2014] Artificial [Qin and Rusu, 2013], [Ahmadi et Gradient Descent immune [Cabanas-Abascal et al., 2013] al., 2014], [Mittal et al., 2014] IEEE Workshop on Large Data Analytics systemsin Transportation Engineering, October 27, 2014 10 / 17
Metaheuristics in Big Data Advantages Adequate for solving NP-Hard problems. Mainly used for: Feature extraction. Dimension reduction. Potential to address multi-objective problems. Challenges The fitness function or the processed data is noisy. The fitness function suffers from approximation errors. October 27, 2014 11 / 17
Section 3 Big Data in Railway Engineering October 27, 2014 12 / 17
Applications Learning to Predict Train Wheel Failures [Yang and Létourneau, 2005] Goal: Optimize maintenance and operation of trains. Approach: Decision Trees and Naïve Bayes. A Simple and Efficient Parallel Approach to Large-Scale Railway Freight Data Analysis [Xie and Li, 2010] Goal: Integrate the national-widely distributed railway freight data sets. Approach: Parallel optimization techniques. October 27, 2014 13 / 17
Applications Improving rail network velocity: A machine learning approach to predictive maintenance [Li et al., 2013] Data from disparate sources: Historical detector data, failure data, maintenance action data, inspection schedule data, train type data and weather data. Approach: Principal Component Analysis, Support Vector Machine. October 27, 2014 14 / 17
Findings Emerging area. Need to integrate data from different sources. October 27, 2014 15 / 17
Section 4 Conclusion and Future Work October 27, 2014 16 / 17
Conclusion Potential to use population-based metaheuristics for Big Data Analytics applied in railway engineering. Opportunity to improve reliability and safety in railway engineering. Future work Identification of the railway data sets. Metaheuristic selection and definition of its use in the Big Data Analytics framework. October 27, 2014 17 / 17
Metaheuristics in Big Data: An Approach to Railway Engineering Silvia Galván Núñez 1,2, and Prof. Nii Attoh-Okine 1,3 1 Department of Civil and Environmental Engineering University of Delaware, Newark, DE, USA 2 silgalnu@udel.edu, 3 okine@udel.edu IEEE Workshop on Large Data Analytics in Transportation Engineering October 27, 2014 October 27, 2014 17 / 17