multi-objective genetic algorithm

Recommender Systems are part of our life nowadays. In almost every big business, recommender systems are suggesting items to users automatically. The only factor that traditional recommender systems take into account is accuracy. They try to estimate user-item ratings as accurate as possible to recommend the more preferable items to users. But from the supplier point of view, profit is the most desirable achievement. In this paper we have proposed a profit-aware recommender system based on the multi-objective genetic algorithm. Our proposed method consider two objectives at the same time: Profit and Accuracy. Profit is amount of profit that company will gain of selling items and accuracy measures how much recommendations are close to user preferences. the NSGA-II as the most successful MO-GA has been selected here and its Crossover and Mutation operations have been designed. Our method and traditional collaborative-filtering method have been implemented and tested on three different datasets: Movielens, Netflix and Yahoo. Results confirm that in both accuracy and net-profit our method prevail over CF method.

Web Services provides a solution to web application integration. Due to the significance of trust in choosing the proper web service, a novel optimal configuration of neural networks by multi-objective genetic algorithm and ensemble-classifier approach is used to evaluate the trust of single web services. For evaluating trust in the single web services, first, a set of neural networks were trained by the settings their parameters through the multi-objective genetic algorithm. Next, the best combination of neural networks was selected to make an ensemble classifier. This method was evaluated with single WS dataset considered eight criteria. Three measurements, accuracy, time and ROC curve were considered to assess the efficiency. Ultimately, the results show that the proposed approach can achieve a trade-off between time and accuracy by the multi-objective genetic algorithm. Also using ensemble-classifiers approach increases the reliability of the model. Consequently, the proposed method promote the detection accuracy.

Handwritten digit recognition, as an important issue in pattern classification, has received considerable attention of many researchers to develop theoretical and practical aspects of this problem. The goal is to recognize a printed digit text or a scanned handwritten using an automated procedure. In this paper an optical character recognition system with a multi-step procedure is presented for Farsi handwritten digit classification. First, a pre-processing course is performed on the image to enhance and make prepare the image for the main steps. Then multiple features are extracted which are believed to be effective in the classification step, among which a multi-objective genetic algorithm selects those with more discriminative characteristics in order to reduce the computational complexity of the classification steps. Following this, only the selected features are extracted from the digit images to be entered to three classifiers. Once the classifiers are trained, their outputs are fed into a classifier ensemble to make the final decision. The weights of the linear combination of classifiers are adjusted by an evolutionary firefly algorithm which optimizes the F-criterion. Evaluation of the proposed technique on the standard HODA database demonstrates that the algorithm of this paper attains higher performance indices including F-measure of 98.88%, compared to other existing methods.

A mobile ad-hoc network (MANET), a set of wirelessly connected sensor nodes, is a dynamic system that executes hop-by-hop routing independently with no external help of any infrastructure. Proper selection of cluster heads can increase the life time of the Ad-hoc network by decreasing the energy consumption. Although different methods have been successfully proposed by researchers to tackle this problem, nearly all of them have the deficiency of providing a single combination of head clusters as the solution. On the contrary, in our proposed method, using a Multi-Objective Genetic Algorithm, a set of near optimum solutions is provided. In the proposed method, energy consumption, number of cluster heads, coverage and degree difference are considered as objectives. Numerical results reveal that the proposed algorithm can find better solutions when compared to conventional methods in this area namely, weighted clustering algorithm (WCA), comprehensive learning particle swarm optimization (CLPSO) and multi objective particle swarm optimization(MOPSO).

Nowadays, developed and developing countries using smart systems to solve their transportation problems. Parking guidance intelligent systems for finding an available parking space, are considered one of the architectural requirements in transportation. In this paper, we present a parking space reservation method based on adaptive neuro-fuzzy system(ANFIS) and multi-objective genetic algorithm. In modeling of this system, final destination, searching time and cost of parking space have been used. Also, we use the vehicle ad-hoc network (VANET) and time series, for traffic flow predict and choose the best path. The benefits of the proposed system are declining searching time, average the walking and travel time. Evaluations have been performed by the MATLAB and we can see that the proposed method makes a good sum of best cost which is useful and meaningful in a parking space reserved for drivers and facility managers. The simulation results show that the performance and accuracy of the method have been significantly improved compared to previous works.

Different approaches have been proposed for feature selection to obtain suitable features subset among all features. These methods search feature space for feature subsets which satisfies some criteria or optimizes several objective functions. The objective functions are divided into two main groups: filter and wrapper methods.  In filter methods, features subsets are selected due to some measures like inter-class distance, features statistical independence or information theoretic measures. Even though, wrapper methods use a classifier to evaluate features subsets by their predictive accuracy (on test data) by statistical resampling or cross-validation. Filter methods usually use only one measure for feature selection that does not necessarily produce the best result. In this paper, we proposed to use the classification error measures besides to filter measures where our classifier is support vector machine (SVM). To this end, we use multi objective genetic algorithm. In this way, one of our feature selection measure is SVM classification error. Another measure is selected between mutual information and Laplacian criteria which indicates informative content and structure preserving property of features, respectively. The evaluation results on the UCI datasets show the efficiency of this method.