MOEA-ARM

Multi Objectives Evolutionary Algorithm - Association rule mining

Introduction

Gather a huge number of multi objectives evolutionary algorithm for solve the association rule mining problem. This library provide state-of-the-art algorithm and tools to compare algorithm performance, futhermore you can choose the metrics which be used for objectives. It's also possible to speed up algorithm by using GPUs.

Experiments

This library is initialy design to be used in experiments for articles, you have then the experiments module allowing you to specified a list of algorithm of criterion a number of repetition and it will perform the experiment and save the results somewhere:

nbIteration = 20
nbRepetition = 5
populationSize = 200
objectiveNames = ['support','confidence','cosine']
criterionList = ['scores','execution time']
algorithmNameList = ['MOCSOARM','MOSSOARM']
perf = Performances(algorithmNameList,criterionList,objectiveNames)
d = Data('Data/Transform/tae.csv',header=None,indexCol=None)
d.ToNumpy()
E = Experiment(algorithmNameList,objectiveNames,criterionList,d.data,populationSize,nbIteration,nbRepetition,path='Experiments/TAE/',display=True)
E.Run()

Hyperparameters managment

It's possible to perform a RandomSearch on the hyperparameters of an algorithm and next save then load latter hyperparameters with the best performances. You can do this using this template :

populationSize = 200
nbIteration = 10,
objectiveNames = ['support','confidence','klosgen']
parameterNames = ['s','a','c','f','e','w']
d = Data('Data/Transform/congress.csv',header=0,indexCol=0)
d.ToNumpy()
modaarm = MODAARM(d.data.shape[1],populationSize,nbIteration,len(objectiveNames),objectiveNames,d.data)
hyper = HyperParameters(parameterNames)
hyper.RandomSearch(30,modaarm,d.data)
hyper.SaveBestParameters('HyperParameters/MODAARM/bestParameters.json')

Performance analysis tools

There is an example of instantiate the performance component.

objectiveNames = ['support','confidence','klosgen']
criterionList = ['scores','execution time','distances','coverages']
algorithmNameList = ['CSOARM','MOPSO']
perf = Performances(algorithmNameList,criterionList,objectiveNames)

There is an example of update the performance component, usually in the main loop

self.perf.UpdatePerformances(score=alg.fitness.paretoFront, executionTime=alg.executionTime, i=i,algorithmName=self.algListNames[k],coverage=alg.fitness.coverage,distance=alg.fitness.averageDistances)

Domination scores graph

graph = Graphs(objectiveNames,perf.scores,path='./Figures/Comparison/paretoFront'+str(i),display=False)
graph.GraphScores()

Execution time graph

Display the execution time for each iteration of each algorithm

graph = Graphs(['execution Time'],perf.executionTime,path='./Figures/Comparison/execution_time')
graph.GraphExecutionTime()

Display the average execution time for the full execution time, that mean the sum of the execution time of each iteration.

g = Graphs(objectiveNames,[],path='../Experiments/RISK/Graphs/ExecutionTime/',display=True,save=True)
g.GraphAverageExecutionTime('../Experiments/RISK/',algorithmNameList,nbIteration)

Number of Rules

This graph allow to know how many rules each algorithm find in his pareto front. There is how compute and display the number of rules.

graph = Graphs(objectiveNames, perf.nbRules, path='./Figures/Comparison/nbRules' + str(i), display=True)
graph.GraphNbRules()

Distances

This graph display the average distance between one rule and all the others. The rules considered are the pareto front's. That allow us to know if the rules are diversified. python g = Graphs(objectiveNames,[],path='../Experiments/RISK/Graphs/Distances/',display=True,save=True) g.GraphAverageDistances('../Experiments/RISK/',algorithmNameList)

Coverage

This graph display the number of row of the dataset cover by the pareto front. python g = Graphs(objectiveNames,[],path='../Experiments/RISK/Graphs/Coverages/',display=True,save=True) g.GraphAverageCoverages('../Experiments/RISK/',algorithmNameList)

Fitness functions

This graphs display the value of each fitness function for each iteration. python g = Graphs(objectiveNames,[],path='../Experiments/RISK/Graphs/LeaderBoard/') g.GraphExperimentation(algorithmNameList,'../Experiments/RISK/','LeaderBoard',nbIteration)

List of available algorithms

NSGAII Non-dominated Sorting Genetic Algorithm II

DEB, Kalyanmoy, PRATAP, Amrit, AGARWAL, Sameer, et al. A fast and elitist multiobjective genetic algorithm: NSGA-II. IEEE transactions on evolutionary computation, 2002, vol. 6, no 2, p. 182-197.

MOWSAARM MultiObjective Wolf Search Algorithm Association Rule Mining

AGBEHADJI, Israel Edem, FONG, Simon, et MILLHAM, Richard. Wolf search algorithm for numeric association rule mining. In : 2016 IEEE International Conference on Cloud Computing and Big Data Analysis (ICCCBDA). IEEE, 2016. p. 146-151.

HMOFAARM Hybrid MultiObjective Firefly Algorithm Association Rule Mining

WANG, Hui, WANG, Wenjun, CUI, Laizhong, et al. A hybrid multi-objective firefly algorithm for big data optimization. Applied Soft Computing, 2018, vol. 69, p. 806-815.

MOCSOARM MultiObjective Cockroach Swarm Optimization Association Rule Mining

KWIECIEŃ, Joanna et PASIEKA, Marek. Cockroach swarm optimization algorithm for travel planning. Entropy, 2017, vol. 19, no 5, p. 213.

MOSAARM MultiObjective Simulated Annealing Association Rule Mining

NASIRI, Mehdi, TAGHAVI, Leyla Sadat, et MINAEE, Behrouz. Multi-Objective Rule Mining Using Simulated Annealing Algorithm. J. Convergence Inf. Technol., 2010, vol. 5, no 1, p. 60-68.

MOBARM MultiObjective Bat Association Rule Mining

HERAGUEMI, Kamel Eddine, KAMEL, Nadjet, et DRIAS, Habiba. Multi-objective bat algorithm for mining interesting association rules. In : International Conference on Mining Intelligence and Knowledge Exploration. Springer, Cham, 2016. p. 13-23.

MOPSO MultiObjective Particle Swarm Optimization

COELLO, CA Coello et LECHUGA, Maximino Salazar. MOPSO: A proposal for multiple objective particle swarm optimization. In : Proceedings of the 2002 Congress on Evolutionary Computation. CEC'02 (Cat. No. 02TH8600). IEEE, 2002. p. 1051-1056.

MOCatSOARM MultiObjective Cat Swarm Optimization Association Rule Mining

BAHRAMI, Mahdi, BOZORG-HADDAD, Omid, et CHU, Xuefeng. Cat swarm optimization (CSO) algorithm. In : Advanced optimization by nature-inspired algorithms. Springer, Singapore, 2018. p. 9-18.

MOTLBOARM MultiObjective Teaching learning Based Optimization Association Rule Mining

SARZAEIM, Parisa, BOZORG-HADDAD, Omid, et CHU, Xuefeng. Teaching-learning-based optimization (TLBO) algorithm. In : Advanced optimization by nature-inspired algorithms. Springer, Singapore, 2018. p. 51-58.

MOFPAARM MultiObjective Flower Pollination Algorithm Association Rule Mining

AZAD, Marzie, BOZORG-HADDAD, Omid, et CHU, Xuefeng. Flower pollination algorithm (FPA). In : Advanced optimization by nature-inspired algorithms. Springer, Singapore, 2018. p. 59-67.

MOALOARM MultiObjective Ant Lion Optimization Association Rule Mining

MANI, Melika, BOZORG-HADDAD, Omid, et CHU, Xuefeng. Ant lion optimizer (ALO) algorithm. In : Advanced optimization by nature-inspired algorithms. Springer, Singapore, 2018. p. 105-116.

MODAARM MultiObjective Dragonfly Algorithm Association Rule Mining

ZOLGHADR-ASLI, Babak, BOZORG-HADDAD, Omid, et CHU, Xuefeng. Dragonfly Algorithm (DA). In : Advanced Optimization by Nature-Inspired Algorithms. Springer, Singapore, 2018. p. 151-159.

MOHBSOTSARM MultiObjective Hybrid Bee Swarm Optimization Tabu Search Association Rule Mining

DJENOURI, Youcef, HABBAS, Zineb, DJENOURI, Djamel, et al. Diversification heuristics in bees swarm optimization for association rules mining. In : Pacific-Asia Conference on Knowledge Discovery and Data Mining. Springer, Cham, 2017. p. 68-78.

MODEARM MultiObjective Differential Evolution Association Rule Mining

ALATAS, Bilal, AKIN, Erhan, et KARCI, Ali. MODENAR: Multi-objective differential evolution algorithm for mining numeric association rules. Applied Soft Computing, 2008, vol. 8, no 1, p. 646-656.

NSHSDEARM Non-Dominated Sorting Harmony Search Differential Evolution Association Rule Mining

YAZDI, Jafar, CHOI, Young Hwan, et KIM, Joong Hoon. Non-dominated sorting harmony search differential evolution (NS-HS-DE): A hybrid algorithm for multi-objective design of water distribution networks. Water, 2017, vol. 9, no 8, p. 587.

MOGEAARM MultiObjective Gradient Evolution Algorithm Association Rule Mining

ABDI-DEHKORDI, Mehri, BOZORG-HADDAD, Omid, et CHU, Xuefeng. Gradient Evolution (GE) Algorithm. In : Advanced Optimization by Nature-Inspired Algorithms. Springer, Singapore, 2018. p. 117-130.

MOGSAARM MultiObjective Gravitational Search Algorithm Association Rule Mining

RASHEDI, Esmat, NEZAMABADI-POUR, Hossein, et SARYAZDI, Saeid. GSA: a gravitational search algorithm. Information sciences, 2009, vol. 179, no 13, p. 2232-2248.

MOSSOARM MultiObjective Social-Spider Optimization Association Rule Mining

CUEVAS, Erik, CIENFUEGOS, Miguel, ZALDÍVAR, Daniel, et al. A swarm optimization algorithm inspired in the behavior of the social-spider. Expert Systems with Applications, 2013, vol. 40, no 16, p. 6374-6384.

MOWOAARM MultiObjective Whale Optimization Algorithm Association Rule Mining

MIRJALILI, Seyedali et LEWIS, Andrew. The whale optimization algorithm. Advances in engineering software, 2016, vol. 95, p. 51-67.

MOSOSARM MultiObjective Symbiotic Organisms Search Association Rule Mining

CHENG, Min-Yuan et PRAYOGO, Doddy. Symbiotic organisms search: a new metaheuristic optimization algorithm. Computers & Structures, 2014, vol. 139, p. 98-112.

MOCSSARM MultiObjective Charged System Search Association Rule Mining

KAVEH, A. et TALATAHARI, Siamak. A novel heuristic optimization method: charged system search. Acta Mechanica, 2010, vol. 213, no 3, p. 267-289.

TheophileBERTELOOT/MOEA-ARM