Search

Abstract: Frequent itemset mining from data streams is an important data mining problem with broad applications such as retail market data analysis, network monitoring, web usage mining, and stock market prediction. However, it is also a difficult problem due to the unbounded, high-speed and continuous characteristics of streaming data. Therefore, extracting frequent itemsets from more recent data can enhance the analysis of stream data. In this paper, we propose an efficient algorithm, called Max-FISM (Maximal-Frequent Itemsets Mining), for mining recent maximal frequent itemsets from a high-speed stream of transactions within a sliding window. According to our algorithm, whenever a new transaction is inserted in the current window only its maximum itemset should be inserted into a prefix tree-based summary data structure called Max-Set for maintaining the number of independent appearance of each transaction in the current window. Finally, the set of recent maximal frequent itemsets is obtained from the current Max-Set. Experimental studies show that the proposed Max-FISM algorithm is highly efficient in terms of memory and time complexity for mining recent maximal frequent itemsets over high-speed data streams. [Copyright &y& Elsevier]

In recent years, classification learning for data streams has become an important and active research topic. A major challenge posed by data streams is that their underlying concepts can change over time, which requires current classifiers to be revised accordingly and timely. To detect concept change, a common method is to observe the online classification accuracy. If accuracy drops below some threshold value, a concept change is deemed to have taken place. An implicit assumption behind this methodology is that any drop in accuracy can be interpreted as a symptom of concept change. Unfortunately however, this assumption is often violated in the real world where data streams carry noise and missing values that can also introduce a significant reduction in classification accuracy. To compound this problem, traditional noise cleansing methods are not applicable to data streams. These methods normally need to scan data multiple times whereas learning in data streams can only afford one-pass scan because of data's high speed and huge volume. To solve these problems, this paper proposes a novel classification algorithm, Class Specific Fuzzy Decision Trees (CSFDT), which utilizes fuzzy logic to classify data streams. The base classifier of CSFDT is a binary fuzzy decision tree. Whenever the problem of concern contains q classes (q > 2), CSFDT learns one binary classifier for each class to distinguish instances of this class from instances of the remaining (q - 1) classes. The CSFDT's advantages are three folds. First, it offers an adaptive structure to effectively and efficiently handle concept change. Second, it is robust to noise. Third, it deals with missing values in an elegant way. As a result, accuracy drop can be safely attributed to concept change. Extensive evaluations are conducted to compare CSFDT with representative existing data stream classification algorithms on a large variety of data. Experimental results suggest that CSFDT provides a significant benefit to data stream classification in real-world scenarios where concept changes, noise and missing values coexist. [ABSTRACT FROM AUTHOR]

The knowledge obtained from emotions via online communities is substantially valuable in various domains, including social management, resource planning, politics, and market predictions. Affective computing, as a multi-aspect realm, aims to exploit emotion-pertinent details from various contents via connecting artificial intelligence to cognitive science. The hidden personality cues in daily brief contents can reveal the cognitive aspect of authors and uncover both similarities and contrasts between them. However, the main challenge lies in devising a cognition-aware algorithm to trace emotional cues in brief contents. To solve the challenge, we develop a novel framework that infers the cognitive aspect of individuals. We propose a deep ensemble method, supplied with a novel dropout algorithm, that aggregates outcomes from various classifiers to extract emotions from short texts. We employ a new embedding approach to enrich emotion-relevant features, collectively assembled via lexicons and attention actuates, resulting in a preferable set of vectors. The experimental results show that our proposed framework can achieve better accuracy in recognizing emotions versus other trending competitors. We empirically observe that detecting emotion latent cues via relying on personality features can effectively distinguish short text authors. Furthermore, the deep learning models overcome conventional methods, including the SVM, categorization, and heuristic rules. [ABSTRACT FROM AUTHOR]

In recent years, with the increasing volume of databases, the removal of redundant features has become an essential thing in classification. A smaller subset of features is selected using feature selection algorithm. One of the famous algorithms of feature selection methods is the crow search algorithm (CSA). This algorithm's popularity can be mentioned in the algorithm's implementation and process and the impressive results compared to the previous algorithms. Despite all these benefits, this algorithm suffers from problems such as unbalanced global and local search. It is also stuck in local optimization due to the search approach's inadequacy. In this paper, a new algorithm based on CSA is introduced. In order to overcome the shortcoming, four fundamental changes have been made to CSA. (i) The algorithm uses the concept of dynamic awareness probability to solve the balance between exploitation and exploration. Then, a new approach is introduced for each part of the search that improves crows' search performance both (ii) locally and (iii) globally. Also, as the last change, (iv) the concept of chaos is used to increase the algorithm's convergence rate. The proposed method has been tested and compared with ten well-known algorithms in this field on the same datasets and has performed on average 20% better in the feature reduction index and 2.5% in the fitness index, while has a lower performance in accuracy by only 1.5%. Practical results show that the algorithm changes have provided attractive results compared to other algorithms in this field in the mentioned metrics. [ABSTRACT FROM AUTHOR]

Feature selection techniques have been presented to allow us to choose a small subset of the original components' relevant features by removing irrelevant or redundant features. Feature selection is essential for many reasons such as simplification, performance, computational efficiency, and quality interpretability. Owing to the importance mentioned above, many researchers have proposed and developed many algorithms to solve the feature selection problem. Although these approaches produce useful results, they possess some shortcomings like inadequate feature reduction. In this paper, a novel feature selection algorithm based on the crow search algorithm is presented. The algorithm uses dynamic awareness probability to keep the balance between the local and global search processes. Moreover, a novel neighborhood assigning strategy has been introduced to optimize the local search. Considering the best-selected features in each iteration helps attain more benefits in global search. The main superiority of the proposed algorithm is the significant feature reduction along with retaining the accuracy. Compared to enhanced crow search algorithm, the proposed algorithm has improved the feature reduction metric and fitness metric by 27.12% and 5.16%, respectively, while losing the accuracy metric by only 0.53%. Several popular UCI datasets have been employed to evaluate the proposed feature selection algorithm. The experimental results show that the proposed algorithm outperformed other feature selection algorithms in state-of-the-art related works regarding feature reduction and accuracy. • Improve the balance between the local and global search. • Introducing a new neighborhood concept for improving the local search. • Proposing a new method to search more purposeful during exploration. • Increasing convergence rate using chaos. • Being pioneer in reducing the dataset volume. [ABSTRACT FROM AUTHOR]

One-versus-all (OVA) classifiers learn κ individual binary classifiers, each distinguishing the instances of a single class from the instances of all other classes. To classify a new instance, the κ classifiers are run, and the one that returns the highest confidence is chosen. Thus, OVA is different from existing data stream classification schemes whose majority use multiclass classifiers, each discriminating among all the classes. This paper advocates some outstanding advantages of OVA for data stream classification. First, there is low error correlation and, hence, high diversity among OVA's component classifiers, which leads to high classification accuracy. Second, OVA is adept at accommodating new class labels that often appear in data streams. However, there also remain many challenges to deploy traditional OVA for classifying data streams. First, traditional OVA does not handle concept change, a key feature of data streams. Second, as every instance is fed to all component classifiers, OVA is known as an inefficient model. Third, OVA's classification accuracy is adversely affected by the imbalanced class distributions in data streams. This paper addresses those key challenges and consequently proposes a new OVA scheme that is adapted for data stream classification. Theoretical analysis and empirical evidence reveal that the adapted OVA can offer faster training, faster updating, and higher classification accuracy than many existing popular data stream classification algorithms. We expect these results to be of interest to researchers and practitioners because they suggest a simple but very elegant and effective alternative to existing classification schemes for data streams. [ABSTRACT FROM AUTHOR]

The purpose of the intrusion detection system (IDS) database is to detect transactions that access data without permission. This paper proposes a novel approach to identifying malicious transactions. The approach concentrates on two aspects of database transactions: (1) dependencies among data items and (2) variations of each individual data item which can be considered as time-series data. The advantages are threefold. First, dependency rules among data items are extended to detect transactions that read or write data without permission. Second, a novel behaviour similarity criterion is introduced to reduce the false positive rate of the detection. Third, time-series anomaly analysis is conducted to pinpoint intrusion transactions that update data items with unexpected pattern. As a result, the proposed approach is able to track normal transactions and detect malicious ones more effectively than existing approaches. [ABSTRACT FROM AUTHOR]