نشریه مدیریت مهندسی و رایانش نرم
سال چهارم شماره 2 (پیاپی 7، پاییز و زمستان 1397)

Nowadays, pervasive environment development has garnered lots of attentions. In such environments, user-object interactions along time are recorded via several sensors, and sensor events are processed as a stream of data. In this process, user’s activities are recognized, and accordingly, essential services are provided. In many activity recognition approaches, firstly the input data stream is segmented, then the activity pertaining to each segment is induced. In such approaches, sensor data stream segmentation is a predominant phase. In this paper, this problem is investigated and a novel method, based on a difference of convex programming problem, is proposed to solve it. In the proposed method a feature vector is calculated for each sensor event in the data stream using a Bayesian approach, and the sequence of such vectors is hired in a difference of convex cost function. The cost function and feature vectors has been calculated by considering heuristics adopting to smart environments. Data segments are extracted by minimizing the cost function. The segmentation purity and conditional entropy have been calculated to measure the performance. Evaluations show that the proposed method has an acceptable performance comparing to some existing approaches.

By using data mining tools in the field of medical diagnosis, some limitations such as the high cost of some tests or their timing will be addressed. In addition, the existence of errors in some experiments has led researchers to be welcomed by categorization methods. In this regard, the present study, based on the combination of clustering and categorization methods, has proposed a new method for the diagnosis of breast cancer. In this operation, the combination is performed using an iterative algorithm and a dependency propagation clustering algorithm. This method produces weights for variables using an innovative algorithm and forms cluster clusters based on the dependency propagation algorithm. Then the number of clusters as a new variable is added to the data, and in the next step, the block algorithm is implemented on the modified dataset containing the main data and the number of clusters. According to the accuracy index, the weights production continues to reach the highest possible precision. According to the numerical experiments conducted in this study, the combination of the dependency emission clustering algorithm with an average accuracy of 36.98 was the most accurate. In addition, the Wilcoxon assumption test confirmed the superiority of the combined neural network compared to other methods.

One of the most common problems with computer networks is the amount of information in these networks. Meanwhile searching and getting inform about content of textual document, as the most widespread forms of information on such networks, is difficult and sometimes impossible. The goal of multi-document textual summarization is to produce a pre-defined length summary from input textual documents while maximizing documents’ content coverage. This paper presents a new approach for textual document summarization based on paraphrasing and textual entailment relations and formulating the problem as an optimization problem. In this approach the sentences of input documents are clustered according to paraphrasing relation and then the entailment score and final score of a fraction of the header sentences of clusters which have the best score according to the user query is calculated. Finally, the optimization problem is solved via greedy and dynamic programming approaches and while selecting the best sentences, the final summary is generated. The results of implementing the proposed system on standard datasets and evaluation via ROUGE system show that the proposed system outperforms the state-of-the-art systems at least by 2.5% in average.

The main purpose of this research is finding major characteristics of clinical signs in the diagnosis of neuropathic disease in patients with chronic long-term pain. This type of disease is caused by various factors such as war, accidents and sports events. In this research, pain questionnaire of Shafa Neuroscience Research Center in Khatam-ol-Anbia Hospital in Tehran is study. By using the deep neural network and the nearest neighbor and the genetic algorithm and the auto encoder, the list of features was obtained with a precision measurement of 75 percentage. The McGill questionnaire was designated as the best effective feature for Neuropathic Pain.

In recent years, machine learning algorithms are widely used for diagnosis and timely treatment of diseases. Moreover, diagnosis of disease on early stages is very effective in improving the disease and in reducing the cost of treatment for the patient. Heart disease is one of the main causes of death in the world. Several studies have been conducted to diagnose of disease and to design an intelligent and efficient system. In this paper, a hybrid algorithm of Whale Optimization Algorithm and simulated annealing are presented to identify the effective factors in the diagnosis of the disease. The support vector machine algorithm is considered for effective classification of the disease. The proposed approach is evaluated using the Cleveland Heart Disease Data Collection in the UCI database. The proposed algorithm has obtained with an accuracy of 87.78% which is able to diagnose of disease with fewer attributes. The results exhibition the superiority of the proposed method which the proposed approach can help physicians to diagnose and to improve disease in the early stages

One of the most important processing steps in the human vision system is the detection of a scene saliency map. Since saliency map can be applied to algorithms such as segmentation, compression and image retrieval, Researchers have focused on providing an efficient model to recognize it. Although a lot of works have been done in this area, the obtained saliency maps are still not satisfying enough. For this purpose, we propose a simple and supervised algorithm to identify the saliency map using a conditional random field (CRF) and saliency cues. In the proposed method, local contrast, center-bias, and backgroundness features have been used for CRF training. Additionally, a new feature based on matrix decomposition has been employed to improve the performance. In the following, CRF has been trained according to the features of 20 images close to the input image. Finally, input image saliency is estimated according to calculated weights in the training phase, input image saliency cues, and ground truths. The proposed method outperforms other methods in terms of algorithm implementation accuracy and speed.