Applications of Soft Computing for the Web

Recommender Systems (RSs) are new types of internet-based software tools, used to provide personalized recommendations to users by handling information overload problem on the Web. Collaborative Filtering (CF), the most known and commonly used recommendation technique in the domain of RSs, generates recommendations toward items which were preferred by other like-minded users in the past. Computing similarity among users efficiently in case of sparse data is the major concern of CF technique. A recent study observed that Context Awareness in CF (CACF) is the next generation of the traditional user–item RSs which provides more accurate and relevant situational recommendations by incorporating contextual ratings given by the user. In this work, we first extend two-dimensional RSs by incorporating contextual information into fuzzy CF user profile (CA-FCF) through contextual rating count approach. Second, we employ genetic algorithm into CACF (GA-CA-FCF) to learn user preferences on individual hybrid user features. By learning the weights on each feature, the user similarity is computed efficiently. We evaluate our approach with Mean Absolute Error (MAE) and coverage performance measures using LDOS-CoMoDa dataset. Experimental results show that our approach has an acceptable improvement in the accuracy with comparisons to classical CF approaches.

This chapter presents an AHP (Analytic hierarchy Process)-based method for multi-criteria decision-making problem involving college selection. The proposed method is evaluated on a sample dataset collected from U. S. news dataset. The U. S. dataset consists of ratings on various aspects of many different colleges from multiple users. We have used eight criteria and eight colleges. The method builds an analytic hierarchy structure of multiple criteria and alternatives to ease the decision-making process. The ranking of each college is based on the overall score considering multiple criteria. The experimental results demonstrate the effectiveness of the use of AHP in college selection process.

The recommender systems are used to facilitate the users with appropriate choices according to their preferences for various online services. Due to the increasing need, various recommendation systems have been developed including recommendation for music, book, movie, etc. The book recommendation technique usually explores the rating of the users for the particular product to recommend it to other users. Instead of utilizing users’ reviews, we have proposed an authorities recommendation approach which exploits ranking of the books by different top-ranked universities. These rankings are aggregated using OWA. Ordered Weighted Aggregation (OWA), a well-known fuzzy averaging operator, is used to aggregate different rankings of the books given by respective universities. The rank of the books is converted into scores using Positional Aggregation based Scoring (PAS) technique. The linguistic quantifiers are applied over these scores and the value of three linguistic quantifiers, ‘at least half’, ‘most’ and ‘as many as possible’, are compared with amazon ranking, evaluated on the basis of ranks explicitly taken from experts. P@10, FPR@10 and Mean Average Precision (MAP) are evaluated. It is evident from the results that quantifier ‘at least half’ outperformed others in the aforementioned performance metric. It is envisaged that the proposed approach will help the research community in designing the recommender systems to explore the relevant books and meet the expectation of the users in a better way.

Recommender systems (RSs) for the web are used to generate recommendations for a set of items that might be of interest to the user. RSs play an increasingly important role for a user in order to deal with information overload problem on the Web. Recent studies demonstrate that incorporation of soft computing techniques into traditional RSs can improve the accuracy of recommendations. This paper, therefore, presents a review of the field of recommendation systems that comprises soft computing approaches besides the typical user-item information used in most of the classical recommender systems. We also provide the classification for each technique, their ability to address the challenges, explain their framework, and discuss possible extensions to further improvement in the recommendation accuracy, which can be served as a roadmap for research in this area.

Researches have shown that people tend to use Twitter both as news media and social network. In addition to sharing their daily activities, thoughts, etc., people also tweet about happenings at the places near them and retweet about major breaking news posted by traditional news media like BBC, NDTV, etc. A large number of tweets about news events are posted every minute. Also, the tweets are itself noisy, redundant, and bursty due to the social nature of Twitter. The Twitter presents all news tweets mixed with other non-news tweets as a list arranged on a timeline, which makes the comprehension of the news difficult as users are not able to browse all unorganized tweets. In this work, we propose to build an application to extractive summarize the news tweets into a flexible, topic-oriented hierarchy. The topic hierarchy can be generated according to time, locations, persons, or events. The flexible topic-oriented hierarchy will allow browsing the tweets according to the user’s interest. The proposed summarization application will present only all the main tweets related to the selected topic on a timeline.

Vast amounts of data from varying fields are available on the Web. For extraction of useful or desired information, one has to dig deep, using effective and efficient means, into these online sources. The quality of data thus extracted depends largely upon the methodology used to search for the desired results. Search engines have been used for long to look for relevant information on the Web but currently, they are limited to a simple relevance-ranking mechanism. The results obtained by keyword searching are vast and beyond the comprehension of a normal human being. The demand for organized search has increased many folds owing to inherent limitations of traditional Web search. In addition to being highly expressive, the search results should present possibly the extracted heterogeneous data in an integrated fashion and in a form that makes its analysis easy. Wrappers have traditionally been used to programmatically extract data of interest from a Web source but they have their own limitations including their inability to distinguish relevant data from the irrelevant. In such situations, fuzzy-based techniques can be used to mine relevant data from various sources on the Web. This chapter presents an overview of Web data extraction from online publication databases and digital libraries. The importance and effectiveness of fuzzy and hybrid string matching techniques for extraction of publications data are outlined in a detailed fashion. The proposed fuzzy data extraction tool and the results obtained from DBLP (Digital Bibliography & Library Project) live dataset of publications from the field of Computer Science are also presented. Toward the end challenges to be addressed and future directions of research are presented and an extensive bibliography is also included.

A decision-making system is an indispensable tool in every industry today. It not only provides relevant solutions but is also a good source of knowledge acquisition from one human by another. This paper discusses a fuzzy logic-based expert system for crop selection which will assist farmer by considering as input the climatic conditions and soil properties prevailing in his region. The system is found to be effective in predicting the correct crop and it provides an exhaustive list of parameters on account of which it can be used as a template to add new rules. The paper concludes by discussing the possibilities of extending this work by developing a full-blown GUI-based software for the deployment in the farming industry.

The vagueness, fuzziness and uncertainty present in medical diagnosis are handled by fuzzy logic formalisms. Medical diagnosis process is simulated using various techniques, some of which are using fuzzy logic. The simulation of differential diagnosis process developed using fuzzy formalisms works in three stages. Initial screening stage accepts symptoms of patient and after computations results into an output as single or multiple diseases. For multiple diseases, this output is taken as input to Stage II which takes another input as history of the patient to give single or multiple diseases. This output and the investigative tests results are input to Stage III which works on Type 1 Fuzzy Inference System and gives single disease output. Development of a desktop application made the system scope limited to one user. The research web application widens the scope of the system thereby increasing utility of the model. The proposed system uses REST (Representational State Transfer) web service architecture having thin UI (User Interface) client serving multiple devices. Input is communicated to web server by client browser; web server in turn forward the request to application server. The application server executes all the complex algorithms to lift the heavy execution part and it shares the result to the client in the form of REST API via web server. The use of three-tier architecture helps to separate the application to divide the processing load on one machine. The client application is lightweight and stateless resulting minimum load on client browser. The use of Single Page Application (SPA) allows different devices and other applications to utilize same API developed. The web application for simulation of differential diagnosis in gynaecology diseases using fuzzy logic is an innovative step, can be used in medical centres at rural areas or can be a teaching aid to medical students or can be an assistant for general physicians.

Clustering as a part of data mining automates the process of collecting similar documents in a single cluster by grouping like ones together. With the help of clusters, we can organize text documents which are similar at a single place and it helps us to group other unknown documents in future to be assigned to one of the known cluster based on the similarity measure. Automatic clustering is usually based on words. In this work, we have used two approaches for clustering using Neutrosophic logic. While using fuzzy logic we take into account only two values; degree of truth and degree of falsity, whereas, in Neutrosophic logic, a new factor called as indeterminacy is also involved. Indeterminacy applies to the situation when for a particular document it is not sure that to which cluster it belongs. The first approach added the indeterminacy factor of Neutrosophic logic to Fuzzy C Means clustering method and modified the formula which calculates the cluster centers and the truth membership of documents toward clusters. The second approach has three phases. First, generate the dataset according to the relative frequency of words in a document. Second, decide seed documents for different clusters with the help of Euclidean distance between different documents. Finally calculate the T, I, and F values for all documents with respect to all clusters. Then decide the cluster for each document on the basis of T, I, and F values.

Web security threat is an intricate and challenging problem. Researchers have been working on the area of web security. However, the security threats related to web applications still exist. Game theoretic models have been applied to tackle diverse network security threats and have been proved to be useful. This paper focuses on using static fuzzy game theory approach to combat web application security threats. In the web security game, the attacker and the security guard can have too many strategies, which will make the problem computationally more complex and time consuming. Hence, fuzzy mathematics can be used to handle the situation. Hence, a Fuzzy game theoretic approach is proposed based on the interactions between the security guard and the ubiquitous attacker. A non-cooperative game is formulated where the attacker tries to access and modify data assets of an organization and the defender tries to protect the resources. The suboptimal Nash equilibrium guarantees that irrespective of the attacker’s strategy the defender’s strategy is optimal.

A significant amount of data gathered by the web-based applications are analyzed to generate more business opportunity and to make critical decisions based on predictability. Automation with decent intelligence makes it possible to model the amount of associated risk and well-assessed mitigation plan. The deviation of actual results from the prediction could be used to improve the prediction quality using evolutionary techniques. Not only the input and the output of such software engines but the process rules also could be best modelled by fuzzy set theory. This chapter focuses on discussing the fuzzy models, their applicability, and ability to provide solutions to such applications. This chapter also gives an overview of the software components as a high-level design that could be used to understand the relationship between the artifacts. The overall discussion also focuses on understanding the nature of business domain criteria which maps to critical elements of decision making in business intelligence in the context of the application of well-established algorithms that work on the fuzzy data set to formulate a decision. During this part of the discussion, we introduce the common deficiencies in a range of decision-making algorithms that may produce the wrong result. This anomaly may creep in because of the presence of the outliers in the decision table leading to affect the steps of the algorithm. We also propose an algorithm to deal with outliers to rectify the result in three well-known methods.

Gravitational search algorithm (GSA) is a new paradigm for optimization that needs to be explored further to show its full potential. The focus of the current work is to address the most promising problems in wireless sensor networks (WSNs) such as cluster head selection and routing using GSA. In a two-tired architecture of WSN, cluster heads (CHs) are overloaded for receiving and aggregating the data packets from member nodes, thereafter, transmitting them to the base station (BS). Therefore, while selecting CHs proper care should be taken to enhance the life of WSNs. After formation of clusters, the data to be transmitted to the BS via intercluster route so that the life of the network is prolonged. In the current study, a new CH selection strategy is developed with an efficient encoding scheme by formulating a novel fitness function based on the residual energy, intra-cluster distance, and CH balancing factor. In addition, a GSA-based routing algorithm is also devised by considering residual energy and distance as parameters to be optimized. The proposed algorithm (GSA-CHSR) is extensively tested with existing techniques on various scenarios of the network to study the performance. The experimental results confirms the superiority and/or competitiveness of GSA-CHSR as compared with some of the well-known existing methods available the literature, such as DHCR, EADC, Hybrid Routing, GA, and PSO.

Routing is regarded as the process that transfers packets from a given source to a given destination with minimum cost, so the routing algorithm has the most specific role of acquiring, organizing and distributing information regarding various network states. So, routing is considered as one of the most important issue for various wireless infrastructures-less networks as most of the Quality of Service (QOS) parameters are related to how efficiently and effectively routes are managed. The traditional approach like Dijkstra’s shortest path algorithm was not able to optimize the QOS issues along with finding the shortest/fittest path. So this chapter utilizes heuristic-based genetic approach to solve and optimize routing-related issues. The proposed and simulated algorithm finds the fittest path from source to destination and optimizes various QOS parameters like hop count, delay, throughput, etc. The proposed heuristic-based approach is compared with traditional Dijkstra’s approach through MATLAB simulator for further validation and affirmation of presented methodology.

Real-world optimization problems in engineering are becoming increasingly complex and require more efficient techniques for their solution. This paper presents a new optimization algorithm, namely variable flight mosquito flying optimization (V-MFO). It mimics the behavior of mosquitoes to find a hole or an irregularity in a mosquito net. It incorporates a variable flying constant and precision movements of the proboscis instead of constant flying and sliding motion of the mosquitoes likewise in simple mosquito flying optimization (MFO). The algorithm was examined for the global minima on diverse types of benchmark functions of diverse dimensions and modality, such as Ackley, Griewank, Rastrigin, Rosenbrock, and Schwefel functions of 5, 10, and 30 dimensions. The results were compared with five established methods, namely genetic algorithm (GA), particle swarm optimization (PSO), seven-spot ladybird optimization (SLO), artificial bees’ colony (ABC), and mosquito flying optimization (MFO). Consequently, this algorithm was found to be efficient, convergent, and accurate.

In this chapter, we summarize the main contributions of different chapters in the book. The book consists of 14 contributed chapters, which are organized in many different parts.