Predicting links is crucial task for determining future links in complex networks across different real-world domains like information networks, social interactions, and technological networks. The link prediction method utilizes graph topological features to locate common neighborhood, yet it overlooks the importance of nodes within the network. In this context, we seek to utilize the importance of node in the network in the link prediction techniques. Centrality metrics measure a node’s relative importance within the network and demonstrates a strong correlation with future links in complex networks. In our study, we propose a novel link prediction measure called Local-Similarity based on Summation of Degree Centrality (CLP). CLP finds similarity scores for node pairs by considering common neighbors and use the centrality scores of these common neighbors in the prediction task. To assess our approach, we compare it with existing methods like Jaccard coefficient, Preferential Attachment, and a recent measure like Keyword Network Link Prediction based on degree centrality. We conduct experiments on four real-world datasets, and CLP shows significant improvements. On average, there’s a 15% improvement in Area Under the Receiver Operating Characteristic (AUROC) compared to existing methods and a 27% improvement over the recent one. Additionally, there’s an average 20 and 23% enhancement in Area Under Precision Recall (AUPR) compared to existing and recent methods. Our experiments highlight the superior performance of the proposed CLP method

Complex networks, represented as graphs, serve as powerful models for understanding real-world systems composed of interacting entities. These networks offer valuable insights into both their structural and dynamic properties. This study concentrates on three fundamental aspects of complex network analysis: centrality, link prediction, and community detection. Centrality focuses on identifying influential nodes within the network, link prediction aims to forecast potential future connections, and community detection uncovers cohesive substructures. Through a thorough review of relevant literature, an exploration of practical applications, and an evaluation of benchmark datasets, this work presents a comprehensive analysis of these critical challenges and assesses the performance of widely utilized algorithms.

Many practical problems and applications are characterized in the form of a network. If the network becomes huge and complex, it becomes very difficult to identify the partitions and the relationships among each of the network’s nodes. As a result, the graph is divided into communities and several community detection methods are proposed to associate those communities. The formation of virtual clusters or communities often occurs in networks due to the likelihood of individuals with similar choices and desires associating with one another. Detecting these communities holds significant benefits across various applications, such as identifying shared research areas in collaboration networks, detecting protein interaction in biological networks and finding like-minded individuals for marketing and suggestions. Numerous community detection algorithms are applied in different domains. This paper gives a brief explanation of existing algorithms and approaches for community detection like Louvain, Kernighan-Lin, Girvan Neuman, Label Propagation and Leiden algorithms as well as discusses various applications of community detection. We have evaluated our comparison with six different datasets namely biocelegans, ca-netscience, usair97, webpolblogs, email-univ and powergrid for comparing the efficiency of the methods. The modularity and conductance scores are used to assess the caliber of the partitioned community. A special emphasis on the comparison of these community detection methods is concerned and how the quality resembles and the time taken for its evaluation. We have evaluated all these algorithms and concluded that Louvain and Leiden community detection algorithms are used for effective community division in terms of its structure and time

Recommender systems recommend products to users. Almost all businesses utilize recommender systems to suggest their products to customers based on the customer's previous actions. The primary inputs for recommendation algorithms are user preferences, product descriptions, and user ratings on products. Content-based recommendations and collaborative filtering are examples of traditional recommendation systems. One of the mathematical models frequently used in collaborative filtering is matrix factorization (MF). This work focuses on discussing five variants of MF namely Matrix Factorization, Probabilistic MF, Non-negative MF, Singular Value Decomposition (SVD), and SVD . We empirically evaluate these MF variants on six benchmark datasets from the domains of movies, tourism, jokes, and e-commerce. MF is the least performing and SVD is the best-performing method among other MF variants in terms of Root Mean Square Error (RMSE)

The problem of recommender system is very popular with myriad available solutions. Recommender systems recommend items to users and help them in narrowing their search from huge amount of options available to the user. In this work, a novel approach for the recommendation problem is proposed by incorporating techniques from the link prediction problem in social networks. The proposed approach models the typical user-item information as a bipartite network, and predicts future links using link prediction measures, in which link prediction would actually mean recommending an item to a user. The standard recommender system methods suffer from the problems of sparsity and scalability. Since link prediction measures involve computations pertaining to local neighborhoods in the network, this approach would lead to a scalable solution to recommendation. In this work, we present top k links that are predicted by link prediction measures as recommendations to the users. Our work initially applies different existing link prediction measures to the recommendation problem by making suitable adaptations. The prime contribution of this work is to propose a recommendation framework routed from link prediction problem in social networks, that effectively utilizes probabilistic measures of link prediction and embed temporal data accessible on existing links. The proposed approach is evaluated on one movie-rating dataset of MovieLens, two product-rating datasets of Epinions & Amazon and one hotel-rating dataset of TripAdvisor. Results show that the link prediction measures based on temporal probabilistic information prove to be more effective in improving the quality of recommendation. Especially, Temporal cooccurrence probability measure improves the area under ROC curve (AUROC) by 10% for MovieLens, 23% for Epinions, 17% for TripAdvisor, 9% for Amazon over standard item-based collaborative filtering method. Similar improved performance is observed in terms of area under Precision-Recall curve (AUPR) as well as Normalized Rank-Score.

In light of the recent COVID-19 epidemic, users are facing growing difficulties in navigating the vast expanse of Internet content to locate relevant information. In this study, we have developed an information extraction mechanism to address users’ inquiries pertaining to COVID-19, catering to a range of depths in response. To accomplish this objective, the CORD-19 dataset, which has been made available by the Allen Institute for AI, is utilized. This dataset comprises 200,000 scholarly articles that pertain to research papers on the topic of coronavirus. These articles have been sourced from many reputable platforms, such as PubMed’s PMC, WHO, bioRxiv, and medRxiv pre-prints. In addition to the aforementioned document corpus, a supplementary list of topics has been furnished, encompassing inquiries pertaining to the infection. Each topic consists of three levels of representations, namely query, question, and story. Inquiry can take on different forms, with query representing a fundamental form, question serving as a more intermediate form, and narrative embodying a more detailed and elaborate type of inquiry. The proposed model uses various word embedding techniques, such as frequency based (Bag-of-words), semantic based (Word2Vec), a hybrid method which combine frequency with semantic (TF–IDF weighted Word2Vec), as well as sequence cum semantic based (BERT) to fabricate vectors for the documents in the corpus, query, question, narrative, and combinations of them. Once vectors have been created, cosine similarity is employed to identify similarities between document vectors and topic vectors. As compared to frequency and semantic models, BERT demonstrates a higher degree of relevance in retrieving documents. with 90% accuracy. The proposed hybrid model, which is the TF–IDF weighted Word2Vec, achieves an accuracy rate of 87%. This is comparable to the average performance of the BERT-Base model demonstrating computational efficiency.

Many real-world problems can be modelled in the form of complex networks. Social networks such as research collaboration networks and facebook, biological neural networks such as human brains, biomedical networks such as drug-target interactions and protein-protein interactions, technological networks such as telephone networks, transportation networks and power grids are a few examples of complex networks. Any complex system with entities and interactions existing between the entities can be modelled as a graph mathematically, with nodes representing entities and edges reflecting interactions. In numerous real-world circumstances, interactions are not confined to pair of entities. Majority of these intricate systems inherently possess hypergraph structures, characterized by interactions that extend beyond pairwise connections. Existing studies often transform complex interactions at a higher level into pairwise interactions and subsequently analyze them. This conversion frequently leads to both the loss of information and the inability to reconstruct the original hypergraph from the transformed network with pairwise interactions. One of the most essential tasks that can be performed on these graphs is Link Prediction (LP), which is the task of predicting future edges (links) in a graph. LP in graphs is well investigated. This article presents a novel methodology for predicting links in hypergraphs. Unlike conventional approaches that transform hypergraphs into graphs with pairwise interactions, the proposed method directly leverages the inherent structure of hypergraphs in predicting future interaction between a pair of nodes. This is motivated by the fact that hypergraphs enable the depiction of intricate higher-order relationships through hyperlinks, enhancing their representation. Their capacity to capture complex structural patterns improves predictive capabilities. Node neighborhoods within hypergraphs offer a comprehensive framework for LP, where hyperlinks simplify interactions between nodes across cliques. We propose a novel method of Link Prediction in Hypergraphs (LPH) to predict interactions within hypergraphs, maintaining their original structure without conversion to graphs, thus preserving information integrity. The proposed approach LPH extends local similarity measures like Common Neighbors, Jaccard Coefficient, Adamic Adar, and Resource Allocation, along with a global measure, Katz index, to hypergraphs. LPH's effectiveness is assessed on six benchmark hyper-networks, employing evaluation metrics such as Area under ROC curve, Precision, and F1-score. The proposed measures of LP on hypergraphs resulted in an average enhancement of 10% in terms of Area under ROC curve compared to contemporary as well as conventional measures. Additionally, there is an average improvement of 70% in precision and around 50% in F1-score. This methodology presents a promising avenue for predicting pairwise interactions within hypergraphs while retaining their inherent structural complexity as well as information integrity.

Link prediction plays a crucial role in identifying future connections within complex networks, facilitating the analysis of network evolution across various domains such as biological networks, social networks, recommender systems, and more. Researchers have proposed various centrality measures, such as degree, clustering coefficient, betweenness, and closeness centralities, to compute similarity scores for predicting links in these networks. These centrality measures leverage both the local and global information of nodes within the network. In this study, we present a novel approach to link prediction using similarity score by utilizing average centrality measures based on local and global centralities, namely Similarity based on Average Degree (Formula presented.), Similarity based on Average Betweenness (Formula presented.), Similarity based on Average Closeness (Formula presented.), and Similarity based on Average Clustering Coefficient (Formula presented.). Our approach involved determining centrality scores for each node, calculating the average centrality for the entire graph, and deriving similarity scores through common neighbors. We then applied centrality scores to these common neighbors and identified nodes with above average centrality. To evaluate our approach, we compared proposed measures with existing local similarity-based link prediction measures, including common neighbors, the Jaccard coefficient, Adamic–Adar, resource allocation, preferential attachment, as well as recent measures like common neighbor and the Centrality-based Parameterized Algorithm (Formula presented.), and keyword network link prediction (Formula presented.). We conducted experiments on four real-world datasets. The proposed similarity scores based on average centralities demonstrate significant improvements. We observed an average enhancement of 24% in terms of Area Under the Receiver Operating Characteristic (AUROC) compared to existing local similarity measures, and a 31% improvement over recent measures. Furthermore, we witnessed an average improvement of 49% and 51% in the Area Under Precision-Recall (AUPR) compared to existing and recent measures. Our comprehensive experiments highlight the superior performance of the proposed method.

Community detection in biological networks is crucial for understanding complex interactions among biological entities. This research focuses on performing community detection using several algorithms such as Kernighan lin bisection algorithm, Louvain algorithm, Girvan Newman algorithm, Fast Greedy algorithm, and Asynchronous fluid community algorithm on various biological datasets. We evaluated the modularity and partition quality for all the communities using all these algorithms separately and did a comparative analysis on the results. Using those results we were able to identify which algorithm is more efficient and scalable in performing the community detection for biological networks.

Recommender systems filter user preferences and surfing history to provide recommendations. These recommendations are used to capture the user interests for making decisions. Based on the interaction of like, and dislikes of the user the decisions are made. We are using the deep learning techniques to enhance the movie recommendations. It has the ability to extract meaningful patterns from large volumes of data. This study uses Artificial Neural Networks (ANN) to learn features from user behavior and movie metadata, and Recurrent Neural Networks (RNN) to capture temporal patterns in user preferences and thereby enhance the recommendation accuracy by considering both short-term and long-term factors. Additionally, Convolutional Neural Networks (CNN) enhance model capabilities by focusing on input data spatial correlations. By combining CNN's ability to extract hierarchical representations of structural and visual aspects into our recommendation system, we intend to improve material knowledge. These techniques play a vital role in providing recommendations by enabling the personalized preferences to the users. The models are trained on diverse datasets using user ratings and viewing history. Model performance on datasets shows decreased mean square and mean absolute error. This research shows how ANN, RNN, and CNN algorithms can provide reliable movie suggestions. © 2024 IEEE.

Two datasets and the most advanced social network analysis (SNA) were used in this paper to analyze drug-induced liver injury (DILI). DILI severity classes, label sections and version details are provided in dataset 1. Dataset 2 divides connectivity patterns according to route of drug administration, exposing in sharp relief the relationship between substances and their methods. In our study, advanced network models are employed to find key compounds including (0, 'vNo-DILI-Concern', 2), and nodes numbers 01, Oral; as well as number zero. These results provide a new perspective on the relationships around drug safety and add to our knowledge of compound relations in terms of DILI. Network analysis and community detection, which reveal hidden patterns that traditional analytical methods might overlook, enhance the interpretability. This cross-disciplinary approach-label sections, administration routes and severity classifications-makes sure full attention is given to matters of DILI. It advances the science of drug safety assessments and itself points to future research into developing more accurate measures of pharmaceutical safety. © 2024 IEEE.

Recommendation systems play an important role in creating personalized content for consumers, improving their overall experiences across several applications. Providing the user with accurate recommendations based on their interests is the recommender system’s primary goal. Collaborative filtering-based recommendations with the help of matrix factorization techniques is very useful in practical uses. Owing to the expanding size of the dataset and as the complexity increases, there arises an issue in delivering accurate recommendations to the users. The efficient functioning of the recommendation system undergoes the scalability challenge in controlling large and varying datasets. This paper introduces an innovative approach by integrating matrix factorization techniques and community detection methods where the scalability in recommendation systems will be addressed. The steps involved in the proposed approach are: 1) The rating matrix is modeled as a bipartite network. 2) Communities are generated from the network. 3) Extract the rating matrices that belong to the communities and apply MF to these matrices in parallel. 4) Merge the predicted rating matrices belonging to the communities and evaluate root mean square error (RMSE), mean square error (MSE), and mean absolute error (MAE). In our paper different matrix factorization approaches like basic MF, NMF, SVD++, and FANMF are taken along with the Louvain community detection method for dividing the communities. The experimental analysis is performed on five different diverse datasets to enhance the quality of the recommendation. To determine the method’s efficiency, the evaluation metrics RMSE, MSE, and MAE are used, and the time required to evaluate the computation is also computed. It is observed in the results that almost 95% of our results are proven effective by getting lower RMSE, MSE, and MAE values. Thus, the main aim of the user will be satisfied in getting accurate recommendations based on the user experiences.

This study combined content-based and collaborative filtering algorithms to create an all-inclusive movie recommendation system. In order to find trends and provide suggestions based on the tastes of comparable users, collaborative filtering analyzes user–item interaction data. Movie qualities are evaluated using content-based filtering in order to suggest related products. Preprocessing techniques like data cleansing, filtering, and text processing are used in the implementation to extract pertinent information and textual elements. While the textual characteristics are converted into numerical representations using methods like literal_eval and CountVectorizer, the metadata contains information on genres, release dates, cast, crew, and keywords. Using the vectorized characteristics of two videos, the cosine similarity between them is computed. Techniques for collaborative and content-based filtering are used to deliver accurate and customized

Influence maximization deals with finding the most influential subset from a given complex network. It is a research problem that can be resourceful for various markets, for instance, the advertising market. This study reviews the dominant algorithms in the field of influence propagation and maximization from a decade.

The ability of machines to recognize and categorize human activities is known as human activity recognition (HAR). Most individuals today are health aware; thus, they use smartphones or smartwatches to track their daily activities to stay healthy. Kaggle held a challenge to classify six human activities using smartphone inertial signals from 30 participants. HAR’s key difficulty is distinguishing human activities using data so they do not overlap. Expert-generated features are visualized using t-SNE, then logistic regression, linear SVM, kernel SVM, and decision trees are used to categorize the six human activities. Deep learning algorithms of LSTM, bidirectional LSTM, RNN, and GRU are also trained using raw time series data. These models are assessed using accuracy, confusion matrix, precision, and recall. Empirical findings demonstrated that the linear support vector machine (SVM) in the realm of machine learning, as well as the gated recurrent unit (GRU) in deep learning, obtained higher accuracy for human activity recognition

Recommendation systems play a vital role in delivering personalized content to users, thereby enhancing their overall experiences across diverse applications. Collaborative filtering based recommendation systems have demonstrated success through the application of matrix factorization techniques. However, the incessant growth in dataset size and complexity presents challenges regarding the scalability of recommendation algorithms. Consequently, addressing these scalability concerns becomes imperative to ensure the seamless functioning of recommendation systems in handling increasingly large and diverse datasets. This research introduces an innovative method that seamlessly integrates matrix factorization techniques and community detection algorithms to effectively tackle the scalability issue in recommendation systems. Through numerous experiments utilizing real-world datasets, the proposed method's efficiency is thoroughly assessed. These compelling findings underscore the method's potential as a promising solution for constructing robust and scalable recommendation systems effectively. Ultimately, the overarching objective is to enhance user experiences by providing personalized and relevant content recommendations that cater to the evolving needs of modern recommendation systems. By optimizing scalability and recommendation accuracy, this innovative approach seeks to elevate the efficacy and user satisfaction of recommendation systems across various domains.

Any real-world entity with entities and interactions between them can be modeled as a complex network. Complex networks are mathematically modeled as graphs with nodes denoting entities and edges(links) depicting the interaction between entities. Many analytical tasks can be performed on such networks. Link prediction (LP) is one of such tasks, that predicts missing/future links in a complex network modeled as graph. Link prediction has potential applications in the domains of biology, ecology, physics, computer science, and many more. Link prediction algorithms can be used to predict future scientific collaborations in a collaborative network, recommend friends/connections in a social network, future interactions in a molecular interaction network. The task of link prediction utilizes information pertaining to the graph such as node-neighborhoods, paths. The main focus of this work is to empirically evaluate the efficacy of a few neighborhood-based measures for link prediction. Complex networks are very huge in size and sparse in nature. Choosing the candidate node pairs for future link prediction is one of the hardest tasks. Majority of the existing methods consider all node pairs absent of an edge to be candidates; compute prediction score and then the node pairs with the highest prediction scores are output as future links. Due to the massive size and sparse nature of complex networks, examining all node pairs results in a large number of false positives. A few existing works select only a subset of node pairs to be candidates for prediction. In this study, a sample of candidates for LP based are chosen based on the hop distance between the nodes. Five similarity-based LP measures are chosen for experimentation. The experimentation on six benchmark datasets from four domains shows that a hop distance of maximum three is optimum for the prediction task.

Approximately 1.6 million individuals use the e-commerce website 'amazon' to buy things from a variety of categories, including food. Reviewing products by consumers who have already purchased them is beneficial to those who are considering doing so, however reviews can be either positive or negative. The buyer finds it difficult to read through such many evaluations before making a purchase, but machine learning ideas and training models make it possible. Our objective is to categorize the reviews based on the attributes that are present in the dataset in order to address issues like these. Redundancy is present in data when it is presented to us in its raw form. So, since evaluations with a score of 3 are regarded as impartial, we delete them along with redundancy. After that, we use the NLP tool kit (a column in the data set) to preprocess the text by removing any stop words (such as in, as, is, on, and punctuation), and we lowercase each letter. The suggested approach renders the text into machine-understandable language using word embedding techniques. Text processing is necessary because customer reviews written in language that is understood by humans cannot be read by machines. The data must be in a machine-readable language in order to apply any classification technique. We separate the data into train and test set after the preprocessing is complete. After the training is complete, we use this model on a test set of data to determine its accuracy. Next, we utilize classification methods like logistic regression and XG Boost to see how accurate our model is. This study's conclusion involves using the model we developed to predict the review based on previous reviews. In this project, we build a model, feed it with existing reviews, apply it to upcoming reviews, and then forecast if the product is good or not. For this work we have taken the data set from Kaggle.

Wearable devices are parts of the essential cost of goods sold (COGS) in the wheel of the Internet of things (IoT), contributing to a potential impact in the finance and banking sectors. There is a need for lightweight cryptography mechanisms for IoT devices because these are resource constraints. This paper introduces a novel approach to an IoT-based micro-payment protocol in a wearable devices environment. This payment model uses an “elliptic curve integrated encryption scheme (ECIES)” to encrypt and decrypt the communicating messages between various entities. The proposed protocol allows the customer to buy the goods using a wearable device and send the mobile application’s confidential payment information. The application creates a secure session between the customer, banks and merchant. The static security analysis and informal security methods indicate that the proposed protocol is withstanding the various security vulnerabilities involved in mobile payments. For logical verification of the correctness of security properties using the formal way of “Burrows-Abadi-Needham (BAN)” logic confirms the proposed protocol’s accuracy. The practical simulation and validation using the Scyther and Tamarin tool ensure that the absence of security attacks of our proposed framework. Finally, the performance analysis based on cryptography features and computational overhead of related approaches specify that the proposed micro-payment protocol for wearable devices is secure and efficient.

Eating right is the most crucial aspect of healthy living. A nutritious, balanced diet keeps our bodies support fight off diseases. Many lifestyle related diseases such as diabetes and thyroid can often be avoided by active living and better nutrition. Having diet related knowledge is essential for all. With this motivation, an ontology related to food domain is discussed and developed in this work. The aim of this work is to create on ontology model in the food domain to help people in getting right recommendation about the food, based on their health conditions if any.

With the rapid development of mobile medical care, medical institutions also have the hidden danger of privacy leakage while sharing personal medical data. Based on the k-Anonymity and l-diversity supervised models, it is proposed to use the classified personalized entropy l-diversity privacy protection model to protect user privacy in a fine-grained manner. By distinguishing solid and weak sensitive attribute values, the constraints on sensitive attributes are improved, and the sensitive information is reduced for the leakage probability of vital information to achieve the safety of medical data sharing. This research offers a customized information entropy l-diversity model and performs experiments to tackle the issues that the information entropy l-diversity model does not discriminate between strong and weak sensitive features. Data analysis and experimental results show that this method can minimize execution time while improving data accuracy and service quality, which is more effective than existing solutions. The limits of solid and weak on sensitive qualities are enhanced, sensitive data are reduced, and the chance of crucial data leakage is lowered, all of which contribute to the security of healthcare data exchange. This research offers a customized information entropy l-diversity model and performs experiments to tackle the issues that the information entropy l-diversity model does not discriminate between strong and weak sensitive features. The scope of this research is that this paper enhances data accuracy while minimizing the algorithm's execution time.

-

Over the previous decade, Cryptocurrency has maintained a steady increase in popularity. The very nature of cryptocurrencies is such that its imperceptible and ungovernable. These qualities intrigue a large number of people to forecast the future value of distinct cryptocurrencies based on their historical price inflation. This research paper assesses and estimates the price projections and volatility of the cryptocurrency named Ethereum (ETH). We accomplish this objective by the use of 4 machine learning algorithms and 3 deep learning techniques to time series analysis of Ethereum (ETH) prices from August 2015 to December 2021 (2315 days). In terms of RMSE, MAE, MSE, and R2 score, deep learning technique LSTM demonstrated superior prediction accuracy when compared to other learning methods.

History shows that, several cloned and fraudulent websites are developed in the World Wide Web to imitate legitimate websites, with the main motive of stealing sensitive important informational and economic resources from web surfers and financial organizations. This is a type of phishing attack, and it has cost the online networking community and all other stakeholders thousands of million Dollars. Hence, efficient counter measures are required to detect phishing URLs accurately. Machine learning algorithms are very popular for all types of data analysis and these algorithms are depicting good results in battling with phishing when we compare with other classic anti-phishing approaches, like cyber security awareness workshops, visualization approaches giving some legal countermeasures to these cyber-attacks. In this research work authors investigated different Machine Learning techniques applicability to identify phishing attacks and distinguishes their pros and cons. Specifically, various types of Machine Learning techniques are applied to reveal diverse approaches which can be used to handle anti-phishing approaches. In this work authors have experimentally compared large number of ML techniques on different phishing datasets by using various metrics. The main focus in this comparison is to showcase advantages and disadvantages of ML predictive models and their actual performance in identifying phishing attacks.

Mining high utility sequential patterns is observed to be a significant research in data mining. Several methods mine the sequential patterns while taking utility values into consideration. The patterns of this type can determine the order in which items were purchased, but not the time interval between them. The time interval among items is important for predicting the most useful real-world circumstances, including retail market basket data analysis, stock market fluctuations, DNA sequence analysis, and so on. There are a very few algorithms for mining sequential patterns those consider both the utility and time interval. However, they assume the same threshold for each item, maintaining the same unit profit. Moreover, with the rapid growth in data, the traditional algorithms cannot handle the big data and are not scalable. To handle this problem, we propose a distributed three phase MapReduce framework that considers multiple utilities and suitable for handling big data. The time constraints are pushed into the algorithm instead of pre-defined intervals. Also, the proposed upper bound minimizes the number of candidate patterns during the mining process. The approach has been tested and the experimental results show its efficiency in terms of run time, memory utilization, and scalability.

A malware is a computer program which causes harm to software. Cybercriminals use malware to gain access to sensitive information that will be exchanged via software infected by it. The important task of protecting a computer system from a malware attack is to identify whether given software is a malware. Tech giants like Microsoft are engaged in developing anti-malware products. Microsoft's anti-malware products are installed on over 160M computers worldwide and examine over 700M computers monthly. This generates huge amount of data points that can be analyzed as potential malware. Microsoft has launched a challenge on coding competition platform Kaggle.com, to predict the probability of a computer system, installed with windows operating system getting affected by a malware, given features of the windows machine. The dataset provided by Microsoft consists of 10,868 instances with 81 features, classified into nine classes. These features correspond to files of type asm (data with assembly language code) as well as binary format. In this work, we build a multi class classification model to classify which class a malware belongs to. We use K-Nearest Neighbors, Logistic Regression, Random Forest Algorithm and XgBoost in a multi class environment. As some of the features are categorical, we use hot encoding to make them suitable to the classifiers. The prediction performance is evaluated using log loss. We analyze the accuracy using only asm features, binary features and finally both. xGBoost provide a better log-loss value of 0.078 when only asm features are considered, a value of 0.048 when only binary features are used and a final log loss of 0.03 when all features are used, over other classifiers.

The problem of mining high utility time interval sequential patterns with multiple utility thresholds in a distributed environment is considered. Mining high utility sequential patterns (HUSP) is an emerging issue and the existing HUSP algorithms can mine the order of items and they do not consider the time interval between the successive items. In real-world applications, time interval patterns provide more useful information than the conventional HUSPs. Recently, we proposed distributed high utility time interval sequential pattern mining (DHUTISP) algorithm using MapReduce in support of the BigData environment. The algorithm has been designed considering a single minimum utility threshold. It is not convincing to use the same utility threshold for all the items in the sequence, which means that all the items are given the same importance. Hence, in this paper, a new distributed framework is proposed to efficiently mine high utility time interval sequential patterns with multiple minimum utility thresholds (DHUTISP-MMU) using the MapReduce approach. The experimental results show that the proposed approach can efficiently mine HUTISPs with multiple minimum utility thresholds.

An information network is modelled as a graph with vertices denoting entities and links depicting connections within them. Heterogeneous Information Network (HIN) contains multiple types of vertices and multiple types of links. There is vast amount of hidden knowledge available in the HINs. Most of the techniques proposed in the literature are focused on homogeneous networks. The same methods are applied for heterogeneous networks by considering homogeneous projections. But this approach leads to information loss. In this paper, major mining tasks applicable for Heterogeneous Information Networks are reviewed.

Life has reached a stage where we cannot live without internet enabled technology. New devices and services are being invented continuously with the evolution of new technologies to improve our day-to-day lifestyle. At the same time, this opens many security vulnerabilities. There is a necessity for following proper security measures. Cybercrime may happen to any device/service at any time with worst ever consequences. In this study, an overview of the concept of cyber security has been presented. The paper first explains what cyber space and cyber security is. Then the costs and impact of cyber security are discussed. The causes of security vulnerabilities in an organization and the challenging factors of protecting an organization from cybercrimes are discussed in brief. Then a few common cyber-attacks and the ways to protect from them are specified. At last, a famous case study of Mirai's attack on a few high-profile victims and the impact is presented.

Cancer is one of the major causes of death in humans. Early diagnosis of genetic mutations that cause cancer tumor growth leads to personalized medicine to the decease and can save the life of majority of patients. With this aim, Kaggle has conducted a competition to classify clinically actionable gene mutations based on clinical evidence and some other features related to gene mutations. The dataset contains 3321 training data points that can be classified into 9 classes. In this work, an attempt is made to classify these data points using K-nearest neighbors (KNN) and linear support vector machines (SVM) in a multi class environment. As the features are categorical, one hot encoding as well as response coding are applied to make them suitable to the classifiers. The prediction performance is evaluated using log loss and KNN has performed better with a log loss value of 1.10 compared to that of SVM 1.24.

-