Journal of Computer and Robotics
Volume:13 Issue: 2, Summer and Autumn 2020

Developing a reliable data mining method is one of the most challenging issues in the features of advanced computer-based systems. Model reliability in depression disorder detection is the determining p-value or confidence limit for accuracy score. In this regard, data mining evaluation metrics provide a path to knowledge discovery and feature extraction is an important process for discovering patterns in data by exploring and modeling big data. The present paper discussed the data mining approach about detection in depression disorder characterized by symptoms such as sadness, feeling empty, anxiety, and sleep symptoms as well as the loss of initiative and interest inactivity. In this survey, a unique dataset containing sensor data collected from patients with depression was used. For each patient, sensor data were measured over several days. In this respect, the represented dataset could be useful for a better understanding of the relationship between depression and motor activity. On the other hand, to overcome the uncertainties raised from wearable sensors (that caused a significant amount of error in similar previous studies using conventional learning methods such as SVM, LR, NB), and also to increase the efficiency and accuracy of the results and to develop a reliable decision-making framework, the evolutionary hybrid machine learning method (fuzzy-genetic algorithm) will be used. The results show the high accuracy of the proposed method compared to other existing methods.

Multiple Sclerosis (MS) is one of the most important diseases of the central nervous system. This disease causes small lesions detectable in Magnetic Resonance Imaging (MRI) images of the patient’s brain. Because of the small size of the lesions, their distribution, and their similarity to some other diseases, the MS diagnosis can be difficult for specialists and may be mistaken. In this paper, we presented a new method based on deep learning for the automatic classification of MRI images. The proposed method is a combinational architecture from transfer learning and wavelet transform (WT). First, WT was applied to the input MRI image, and its four output sub-bands are used as the input of four fine-tuning networks based on EfficientNet-B3. Transfer learning networks perform feature extraction on all four sub-bands. Then, their outputs are combined, and the result is classified by a fully connected neural network. Due to the feature of WT to extract local features, it was possible to highlight the lesions in the images and subsequently classify it with higher accuracy and precision. Various criteria have been used to evaluate the proposed method. The results of the experiments show that the Values of accuracy, precision, sensitivity, and specificity are 98.91%, 99.20%, 99.20%, and 98.33%, respectively.

High efficiency video coding (HEVC) standard is highly contributive in data hiding. Combining the HEVC standard with data hiding methods is a complex task, and because the highly efficient coding process is a powerful attack, eliminating some of the original video data would not have a significant effect on obtaining the best compression rates and transmission efficiency. The data embedded in the pixels is lost when compressed with the HEVC standard. In order to solve this problem, we need to use features other than the pixel value .In this paper, an efficient data hiding method is proposed through intra prediction modes in HEVC, where the intra prediction modes is applied as the secret data carrier obtained from the N smallest prediction units. The experimental results show that the average PSNR decreased by 0.11 db, while the bit rate increased by on average 0.25%.

Today, with the rise of cloud data centers, power consumption has increased and cloud infrastructure management has become more complex. On the other hand, meeting the needs of cloud users is an important goal in the cloud infrastructure. To solve such problems, an autonomous model with predictive capability is needed to do virtual machine consolidation at runtime effectively. In fact, using the feedback system of autonomous systems can make this process simpler and more optimized. The goal of this research is to propose a cloud resource management model that makes the virtual machine consolidation process autonomous, and by using a prediction method compromises between service level agreement violations and energy consumption reduction. In this research, an autonomous model is presented which detects overloaded servers in the analysis phase by a prediction algorithm. Also, at the planning phase, a multi heuristic algorithm based on learning automata is proposed to find proper servers for virtual machine placement. Cloudsim version 3.0.3 was used to evaluate the proposed model. The results show that the proposed model has reduced averagely the service level agreement violations, energy and migration counts by 67.08%, 11.61% and 70.64% respectively, compared to other methods.

Ensemble learning is one of the learning methods to create a strong classifier through the integration of basic classifiers that includes the benefits of all of them. Meanwhile, weighting classifiers in the ensemble learning approach is a major challenge. This challenge arises from the fact that in ensemble learning all constructor classifiers are considered to be at the same level of distinguishing ability. While in different problem situations and especially in dynamic environments, the performance of base learners is affected by the problem space and data behavior. The solutions that have been presented in the subject literature assumed that problem space condition is permanent and static. While for each entry in real, the situation has changed and a completely dynamic environment is created. In this paper, a method based on the reinforcement learning idea is proposed to modify the weight of the base learners in the ensemble according to problem space dynamically. The proposed method is based on receiving feedback from the environment and therefore can adapt to the problem space. In the proposed method, learning automata is used to receive feedback from the environment and perform appropriate actions. Sentiment analysis has been selected as a case study to evaluate the proposed method. The diversity of data behavior in sentiment analysis is very high and it creates an environment with dynamic data behavior. The results of the evaluation on six different datasets and the ranking of different values of learning automata parameters reveal a significant difference between the efficiency of the proposed method and the ensemble learning literature.

 We present IMPTCHA, a new CAPTCHA (Completely Automated Public Turing test to tell Computers and Humans Apart) as a security measure to recognize human users. The proposed system uses images instead of distorted text to label images as a valuable output. IMPTCHA is generated from images on the Web. For passing this CAPTCHA, users must type two words for description of two images. When users pass the challenge, the provided meaningful labels are used to determine the content of images. In addition, semantic graphs for labels and images are created and according of it we’ll able to develop an image semantic search engine. Due to usage of images in this system, and its architecture, it is highly secure compared to its counterparts. In a user study involving 60 participants, IMPTCHA’s word accuracy is measured to be 98.18% while 61.26% of users could pass the challenge. Keywords:CAPTCHA, Ontology, Image labeling, Security