Sharif Digital Repository

Nowadays, living in information era has forced us to face with a great deal of problems of which the input data is received like a nonstop endless stream. Intrusion detection in networks or filtering spam emails out of legal ones are instances of such problems. In such areas, traditional classification algorithms show function improperly, thus it is necessary to make use of novel algorithms that can tackle these problems. Among classification algorithms, decision trees have significant advantages such as being independent of any parameter and acting robust against outliers or unrelated attributes. Moreover, results of a decision tree are quite easy to interpret and extract rules from.... 

It is not possible to compute a linear classifier to classify real world images, which are the focus of this thesis. Therefore, the space of such images is considered as a complex. In such cases, kernel trick in which data samples are implicitly mapped to a higher dimension space, leads to a more accurate classifier in such spaces. In kernel learning methods, the best kernel is trained for the classification problem in hand. Multiple Kernel Learning is a framework which uses weighted sum of multiple kernels. This framework achieves good accuracy in image classification since it allows describing images via various features. In the image input space which is composed of different extracted... 

Traditional methods for data classification only make use of the labeled data. However, in most of the applications, labeling the unlabeled data is expensive, time consuming and requires expert knowledge. To overcome these problems, Semi-supervised Learning (SSL) methods have become an area of recent research that aim to effectively addressing the problem of limited labeled data.One of the recently introduced SSL methods is the classification based on geometric structure of the data, namely the data manifold. In this approach unlabeled data is utilized to recover the underlying structure of the data. The common assumption is that despite of being represented in a high dimensional space, data... 

Supervised kernel learning has been the focus of research in recent years. Although these methods are developed based on rigorous frameworks, they fail to improve the classification accuracy in real world applications. In order to find the origin of this problem, it should be noted that the kernel function represents a prior knowledge on the labeling function. Similar to other learning problem, learning this prior knowledge needs another prior knowledge. In supervised kernel learning, only naive assumptions can be used as the prior knowledge. These include minimizing the ℓ1 and ℓ2 norms of the kernel parameters.
As an alternative approach, in Semi-Supervised Learning (SSL), unlabeled... 

Trying both to understand the brain and to emulate some of its strengths has been one of the greatest human desires since ancient times. One of these amazing abilities is recognizing via vision. As a result, image recognition has been turned into one of the most attractive areas of research in Computer Vision field since recently. The challenging problem begins to rise where occlusion, scale, rotation and various light conditions contribute and manipulate the paradigm of image recognition. Although recognitions with these challenging problems are some of the capabilities that the brain has, but these are not all. One of the remarkable abilities of the brain is to recognize concepts through... 

One of the problems in signal processing is supervised classification. In supervised classification, the goal is to learn the structures and patterns of a dataset using a set of labeled data called the training dataset to correctly classify data samples that are not used in the training data but follow the same pattern and structure. One approach to this problem that has recently received attention is neural networks. Although this approach has good performance in applications, in order to perform well, they require a large amount of data and many trainable parameters, which result in high computational complexity. Another approach to this problem is dictionary learning-based classification.... 

Image classification, i.e. the task of assigning an image to a class chosen from a predefined set of classes, has addressed in this thesis. At first the classifier is divided into two major sub partitions, feature extraction and classifier. Then we show that by using local feature extraction techniques such as BOW the classification accuracy will improve. In addition, using unlabeled data is argued as the fact to deal with high nonlinear structure of features. Recently, many SSL methods have been developed based on the manifold assumption in a batch mode. However, when data arrive sequentially and in large quantities, both computation and storage limitations become a bottleneck. So in large... 

Distance Metric Learning algorithms have been widely used in Machine Learning methods recently. In these algorithms a distance function between objecs (data points) is learned based on their labels or similarity and dissimilarity constraints. Recent works have shown that a good precision is obtained in classification or clustering methods which use these functions. Since in the current systems many of data points do not exist at the beginning and are added to the training set as the algorithm is run, online methods are needed to update learned metric due to new data.
In this thesis, we proposed a new online distance metric learning method that has higher performance than existing... 

Graphs are powerful tools for modeling real-world data. Data analysis using graphs allows us to study the samples relations and identify rich patterns. Although graph modeling can result in a better understanding of data, it requires having strong methods. Graph neural network models have attracted more attention in recent years. These networks are able to simultaneously analyze data features and their relationships to each other and find node representation in low-dimensional feature space. However, due to the novelty of this field, many challenges are still unexplored.In this study, we intend to examine the challenges in this field by focusing on improving the representation of nodes with... 

Nowadays, researchers working on classifiers which are designed to predict minority class. In this work, we attempt to improve fraud detection performance, with minimum possible complexity. In this regard, by incrementing model sensitivity to minority class samples, we solve the problem of model ignorance to these instances. Moreover, by using clustering, we cluster similar inputs based on their features, and split each class to smaller bins. Then with considering the fact that, prediction probability threshold influences the final performance, we define statistical hypothesis testing exclusively for each cluster to evaluate predictions with expected range. In this method, model is not... 

Deontic logic is used to formalize legal reasoning. To apply this logic in law, we describe tersely some efforts to improve this logic by relativizing its operations with respect to different people and groups of society. Until now, this logic was restricted to formalize “what must be”. We extend this logic to dyadic logic to formalize “what must be done”.
In practice, legal reasoning leads to non-monotonic logics, the most applicable one in law is defeasible logic. So it is necessary to combine deontic and defeasible logics to formalize legal reasoning in a more appropriate way. To do that, we must adjust possible worlds of these two logics. In this way, we find a method for... 

Gas leak explosion is major risks in closed environment in the chemical industry. One of the fundamental principles of safety is control of ignition sources using special protective equipment in areas where possible flammable gases or vapors are there. Computational fluid dynamics is used in this project to evaluate and analyze the risks associated with gas distribution. Ventilation effect has been studied by computational fluid dynamics on gas cloud formed by leakage of flammable in hazardous areas. To demonstrate the accuracy of calculations performed by computational fluid dynamics, the results are validated by experimental data. The simulated results for gas dispersion are found to be in... 

The classification of different tumor types is of great importance in cancer diagnosis and drug discovery. However, most previous cancer classification studies are clinical based and have limited diagnostic ability. Cancer classification using gene expression data is known to contain the keys for addressing the fundamental problems relating to cancer diagnosis. The recent advent of DNA microarray technique has made simultaneous monitoring of thousands of gene expressions possible. With this abundance of gene expression data, researchers have started to explore the possibilities of cancer classification using gene expression data and quite a number of Pattern Recognition approaches have been... 

Developments in the field of data storage which is due to computers have led to an extraordinary increase in medical data just like the increase in all other fields. As a result, physicians are faced with the problem of using the stored data. Therefore, the traditional manual data analysis is inadequate due to the large amounts of data. Furthermore, the ability to use this data to extract useful information is critical for the quality of medical care. Therefore, data mining techniques arose so that we will be able to extract knowledge through applying them to the raw data and subsequently help the doctors in making decisions.
In this study, we are pursuing four goals. First, in order to... 

Data mining is the one of top ten developing knowledge in the world. This study followed three fold objectives; Firstly, An efficient model based on data mining algorithms is proposed to predict the duration of hospitalization time for patients of digestive system disease that need short term care. Duration of hospitalization is an important criterion to be used for predicting the hospital resources. In order to, a combined model based on CHAID and C.5 decision trees and a neural network is suggested. The suggested model predict the duration of hospitalization with 82% accuracy. The second object of this study is to propose an algorithm based on likelihood ratio. The suggested algorithm... 

Nowadays, data is the most important asset for health organizations in which the process of collecting, storing and analyzing of data leads to success of health organizations. Many companies have turned to data mining for the beneficial use of these data. The main purpose of data mining is to obtain useful knowledge from existing data. One of the diseases that is very significant for data miners is cardiovascular disease. Cardiovascular disease is the most important cause of death in the world. Therefore, it is necessary to improve the diagnostic and predictive measures of these patients. In this study, a database containing of characteristics of patients with chest pain who referred to... 

With the growth of social networks, link prediction has attracted great attention. Completing partially observed networks, recognizing errors in observed links, predicting the network’s future structure to aid decision making, and presenting users with favorable links are some the motivations that have made link prediction important and effective for complex networks. In this work, we analyze link prediction in DBLP’s author network and attempt to increase the accuracy of state-of-the-art link prediction techniques by extracting discriminative information from the available metadata. Abstracts are an important resource that indicate an author’s field of study. Extracting the concepts an... 

Recently, data scientists have been paying much attention to sports. Many researches have been done in this field, using data mining and machine learning techniques. The following research aims to predict the results of football matches, which consists of two general approaches. For the first and second approaches, we used video game data and match statistics, respectively. In both approaches, it was tried to predict not only the final result (win, draw, or loss) but also the final goal difference. In the first approach, the home team victory was predicted by 73% accuracy, the draw by 75.4%, and the home team defeat by 73.7%. Nevertheless, in the second approach, the home team victory was... 

Despite the vulnerability of non-structural systems in buildings to disasters, their resilient design has received minimal attention from the practitioners of the construction industry. However, interruption in the performance of these systems jeopardizes the functionality of the buildings and threatens the resilience of the whole community. To address this issue, the present study proposes a novel framework for predicting possible damage states of non-structural building systems under disasters. The proposed framework benefits from an automated combination of Building Information Modeling (BIM) as a visualized 3D database of the building's components and the Machine Learning (ML)... 

One of the challenging issues in machine learning is the problem of data imbalance. Data imbalance occurs when the number of samples from one or more classes significantly exceeds or falls behind that of other classes. The existence of data imbalance in a dataset often leads to misleading accuracy of models, inadequate prediction of minority class, and a lack of generalization. Data imbalance can be observed in various datasets such as fraud detection, disease diagnosis, email spam detection, and fake news detection. To address this issue, various methods have been proposed, categorized into four groups: data-level, algorithm-level, cost-sensitive, and ensemble approaches. In this study, two...