clustering

Time series-clustering, defined as deriving trends and archetypes from sequential data, divides time series into groups considering their characteristics. Previous works mainly focused on distance criterion and clustering algorithm to cluster the time series so few researchers have investigated the similarities between the segments of a time series. To address this research gap, we propose a new two-step approach based on sub-time series and combination clustering. In the rst step, a time series data set is segmented using a xed window size, and each segment is clustered by applying a hierarchical clustering algorithm and Euclidean distance. Also, we use a logarithmic relation based on the length of the time series data set to determine the number of components, selecting the best outcomes using various internal criteria including intergroup variance, Kalinsky-Harbaz, and Dunn index. In the second step, the results of the rst stage are processed using ensemble clustering, and the nal clustering label is obtained. We develop two novel algorithms based on dierent internal criteria for selecting the best segmentations: the rst one in which we consider only one internal criterion and the second one in which we consider three internal criteria simultaneously. Moreover, we run various settings on 82 datasets with 10 replications for the two presented algorithms, checking the nal precision using an external RAND index. Then, to identify the best settings for the proposed algorithms we applied Wilkinson statistical test. Statistical comparison of the results of the two new algorithms on 82 data sets with some algorithms in the related literature indicates signicant improvement In terms of error rate and execution time. Finally, the ndings acquired based on the best settings of the proposed algorithms indicate that the suggested method has the best RAND index among the previous algorithms in the literature for 32% of the dataset tiers.

Pattern recognition is a branch of machine learning that recognizes the patterns and regularities in a set of data, and digit recognition is considered one of the pattern recognition categories. Due to the similarities between some digits in each language, especially in Persian, different algorithms have been developed to recognize the handwriting digits with the least error and in the shortest time complexity. One of the most commonly used methods in data classication is the neural network algorithm. While neural networks have been used in the literature for handwriting digits recognition, the combination of clustering approaches and neural network classiers has not been considered for this problem. Accordingly, this paper proposes an algorithm based on the combination of clustering approaches and neural network classiers to recognize the Persian handwritten digits accurately. This algorithm performs pattern training and recognition based on Probabilistic Neural Networks (PNN) and multilayer perceptron (MLP) neural networks. In this regard, after extracting the characteristic loci feature and zoning from each image in the training database, the data of each of the ten classes has been clustered using linkage, Partition Around Medoids (PAM), and Fuzzy C-Means (FCM) methods based on the extracted features. Then, the new ten classes resulting from the clustering algorithm are taught by one of the two classiers, including MLP and PNN. In order to determine the optimal number of clusters in each class, the Tabu search optimization algorithm, one of the most accurate meta-heuristic optimization algorithms, is used. The performance of the proposed algorithms is evaluated and compared with existing algorithms based on the HODA dataset. Based on the results, the proposed algorithm accurately recognizes the Persian handwritten digits. In addition, the proposed method performs more accurately and much faster than most competing algorithms.

Clustering algorithms are useful tools for understanding data structure and classifying them into different data sets. Due to the importance of using these algorithms in analyzing financial market data that have a high volume and scope, this study in order to select the best clustering algorithm for clustering companies listed on the Tehran Stock Exchange in the field of finance from It has used different clustering algorithms and evaluated the validity of these algorithms and selected the best algorithm.

This research is applied in terms of purpose and descriptive in terms of implementation method and is of quantitative type (mathematical modeling). The statistical population of the research includes 403 companies listed on the Tehran Stock Exchange in 2019, whose performance has been evaluated based on four financial criteria.

After clustering the surveyed companies by five clustering algorithms, namely K-means, EM, COBWEB, density-based algorithm and ward method, seven indicators RS, DB, Dun, SD, Purity, Entropy and Time were used to evaluate the algorithms. Finally, the total performance of the algorithms was analyzed based on TOPSIS, VICOR and DEA methods. Based on the results, K-means has a better performance in clustering based on the financial data sets.

Originality/Value:

 Since no clustering algorithm can have the best performance in all measurements for each data set, this study uses a combination of multiple criteria to analyze data clustering algorithms related to the field of financial performance appraisal. Companies have provided suggestions and the results of this study have been used effectively for investors in the field of finance, which leads to the optimal choice of investment portfolio.

Gaining a competitive advantage is very important for mobile operators. Mobile value-added services are one of the innovations that operators use to diversify their business. Cross-selling is crucial for mobile operators to generate revenue and profits. Because operators will incur lower ancillary costs compared to attracting new customers. But it is not easy for them to identify potential customers who buy the services provided by operators. In this article, an attempt has been made to facilitate the cross-selling of mobile value-added services. The data used in this research is information about the past purchases of the customers of HamrahAval Company from the value-added mobile services. In the proposed solution, the infrastructure for creating cross-selling customer profiles is discussed. In this solution, after determining the optimal category of customers using their clustering, an attempt has been made to discover the rules between the services used by customers. By creating this profile, a target community can be achieved for the cross-selling of each service.

Wireless sensor network is a growing technology.  In wireless sensor networks, performance is usually affected by energy constraints. In this paper, a method is proposed based on the label propagation algorithm for this limitation. At first, the sensors are composed of a graph In the next stage weighing the edges of this graph is based on four similarity measure  Then for each node, the centrality and the initial label are obtained And finally, by updating the lebel,nodes with the same label are placed in a cluster The results of the proposed method have been compared with  the number of live nodes and the mean energy of live nodes measures by the leach method The results indicate that in the proposed method of positioning the sensors and setting the threshold value for the formation of the graph from the sensors are only fundamental variables And the comparison shows that the proposed method is superior to the leach method

In this research, we evaluate energy efficient clustering methods in wireless sensor networks. The results of this study indicate that clustering by DBSCAN has a higher efficiency score than other clustering methods. The DBSCAN scored 99%, but the K-Means algorithm scored 76%. Also, the energy remaining in the network after the simulation is completed in routing with the new protocol is about 11% more than routing with EEHC and about 9% more than routing with LCA protocol. If we consider the lifespan of the network when the first node in the network is turned off, in the new protocol, the first node shuts down 6 seconds later than the EEHC protocol and 12 seconds later than the LCA protocol. This means that on average, about 10% of network life has been increased with the new protocol. The I-LEACH protocol improves energy efficiency and longevity by working more in the same structures than the conventional LEACH protocol.

Today, the use of data mining techniques such as classification, clustering, discover repetitive pattern and discover outliers in different domains including production, medicine, social, meteorology, stock exchange, sales, customer service and other areas are increasing. Data mining techniques are specifically designed for static data. Therefore, their use for time series data requires some modifications to their respective algorithms. One of these changes is the selection of the appropriate similarity measurement method, because similarity measurement methods are used in all data mining techniques. Therefore, in this research, we will evaluate and compare the effect of two commonly used and efficient methods of time series similarity measurement in data mining. This evaluation is done in relation to the effectiveness of these methods in achieving better results. These methods are the Longest Common Sub Sequence (LCSS) method and the Dynamic time Warping (DTW) method. The main purpose of this research is to compare the performance of these methods in time series data mining. The data mining techniques that used in this research are the nearest-neighbor technique and k-medoids clustering algorithm. The performance evaluation process is described in the text. This process uses the nearest-neighbor technique to calculate the accuracy of detection of right time series class, and uses the k-medoids clustering technique to calculate the clustering accuracy, the ability to correctly determine the number of clusters, and the ability to determine the better cluster representative. For this purpose, we use 63 time series data sets by random from a world-renowned database that named UCR collection. The results show that the effect of LCSS method is significantly better than the effect of DTW method on the correct detection accuracy of time series class and clustering accuracy by 99% and 92.5% confidence, respectively, but there is no significant difference between them in terms of their effect in determining the number of clusters and cluster representatives. The results of this research help to use these methods in appropriate data mining techniques in issues such as customer segmentation, workshop scheduling and the like more accurately.

In this research, a new method is presented to improve the recommendation systems in the field of health tourism, which can make accurate predictions by using participatory filtering and by using the points that previous tourists have given to places and health professionals in our country. For the use of tourists. According to the research, data clustering using DBSCAN algorithm obtained 99% efficiency score, which is the highest efficiency score among the existing algorithms. Also, SVM method has 95% score in accuracy section and 99% score in call section. Which shows the high accuracy of predicting the results and the proposed method in general up to 80% can correctly identify the places needed by the tourist and suggest the appropriate place to a large extent correctly