In this study, we develop an innovative federated framework for erasable itemset mining to address the challenges of horizontal federated learning in data mining and resolve the shortcomings of the previous algorithm. The framework is based on a client-server architecture, in which clients from multiple data sources collaborate to effectively integrate information. The proposed algorithm is divided into two parts, including the client-side mining and the server-side aggregation. During the client-side mining stage, the algorithm introduces quasi-erasable itemsets to collect additional useful information, facilitating the integration of results on the server side. In the server-side aggregation stage, the algorithm employs a boundary strategy, effectively utilizing the clients' quasi-erasable and erasable itemsets to improve the accuracy of the results. Experimental results demonstrate that the proposed method not only effectively mines complete knowledge but also ensures data-privacy protection.

In the paper, we propose a two-stage face-inpainting approach based on conditional generative adversarial networks. In the first stage, a deep-learning model is trained for predicting face landmarks. It also dynamically adjusts the penalty value of the loss function based on the view-degree of a face to improve the ability of predicting high view-degree faces. In the second stage, masked face images and their corresponding face landmarks are concatenated to form the condition of training a conditional Generative Adversarial Network (GAN) for inpainting the masked face. If an input masked image is a nearly-frontal face, an additional procedure for face symmetry processing will be performed before the image is input into the inpainting model. The experimental results show that the proposed training method in the first stage can effectively enhance the robustness of the face-landmark prediction model and reduce the impact of data imbalance, thereby improving the effect of later face inpainting. They also show that the proposed face-inpainting model in the second stage can better maintain the geometric structures and symmetric outlooks of inpainted faces than previous ones.

High-utility-itemset mining, extended from frequent-itemset mining, considers external utilities of items in quantitative databases to obtain the itemsets with high utility. To easily understand those patterns with high utility values, fuzzy utility mining adopts fuzzy sets to increase the readability of the derived utility itemsets. However, real-world databases are usually dynamic. New transactions may be intermittently added, and the corresponding mined patterns must be updated to keep correct knowledge. In this paper, we propose an incremental method based on our previously proposed batch-processing fuzzy utility tree-based mining algorithm. The proposed approach adopts the fast-update (FUP) strategy. It considers newly coming data to readjust the head table and the tree structure, which generate desired itemsets in two phases. The experimental results reveal that the proposed algorithm outperforms the tree-based batch mining method.

Erasable itemset mining has been a valuable mining problem for manufacturers. It can extract less profitable materials from a product dataset and provide managers with good decision-making and a trade-off between cost and profit. However, the traditional erasable itemset mining methods seldom consider the time factor. For time-sensitive industries such as agro-processors, the time range is important in determining which materials are less profitable. Hong et al. first proposed the concept of temporal erasable itemset mining and seven lifespan options. They also proposed a unified temporal erasable (UTE) mining algorithm for getting incomplete temporal erasable itemsets. Howerever, the UTE algorithm does not satisfy the property of downward closure. In this work, we propose an improved algorithm to improve the performance of the UTE algorithm based on a lower-bound strategy and satisfying the property of downward closure. The proposed algorithm uses a hash table to store information that will be reused during the mining process to avoid scanning a dataset multiple times. The designed lower-bound strategy can preserve the downward closure property, narrowing the search space of candidate itemsets. In numerical experiments, we compare the performance using several metrics between the proposed method and the previous work. From the results of experiments, our proposed method outperforms the existing method on various metrics, such as execution time and the number of candidate erasable itemsets.

In this paper, we propose an automatic extraction process of the dimension parameters shown in three-view drawings. It is divided into two stages. In the first stage, we detect three-view drawings in datasheets and find out the text regions containing the parameters in the drawings by deep learning. We then recognize the values in these regions. In the second stage, we design two algorithms, based on k-nearest neighbors (k-NN) and statistical evaluation, respectively, to match the digitized parameters with the values. We also conduct experiments to show the high accuracy in the two stages.

Recently, many image recognition applications employing convolutional neural networks have great success with satisfactory accuracy performances due to the rapid development of computer hardware and neural network technology. In these applications, convolutional neural networks used a group of rotated images in random angles from the same object as the network inputs during training to effectively recognize rotated objects. However, as shown in this study, this rotated-image approach could not effectively learn rotational features. In this paper, we present the multi-scale rotation-invariant features used in the convolutional neural network to identify rotational invariance. It can successfully capture rotational features for various data sets. Our model is established through the concepts of dihedral group transformations, multi-level, multi-scale, rotation-invariant pooling, and sharing weights of the convolutional neural network to learn convolution kernels that can achieve rotational invariance. Our model showed a significant improvement over other ones and practically learned the rotational invariance of an object.

During our research conducted in this paper, we demonstrate a successful mining progress to mine the sequential high-utility patterns of uncertain databases. A PUL-Chain structure is developed and built in this paper with several pruning methods to decrease the search space of required patterns for mining efficiency improvement. In contrast to the standard HUS-Span, our experimental results show clearly that both in runtime as well as in the number of candidates discovered, the developed algorithms showed the effectiveness of the discovered patterns and its mining efficiency compared to the elder HUS-Span model. We present the details of our research here in this paper and also focus our attention to future directions that this research may take in the years to come.

More complicated than fuzzy data mining, temporal fuzzy utility data mining takes into account the temporal factor of transactions, purchased quantities, item profits, and linguistic terms. In this paper, a tree structure modified from the frequent-pattern tree is designed and a mining algorithm based on it was proposed to extract high temporal fuzzy utility patterns from transactional datasets with the temporal property. The method requires two-phase processing to find all high temporal fuzzy utility itemsets. Experimental results show that the proposed algorithm performs better than the Apriori-based mining algorithm.

Comparing to frequent itemset mining (FIM), utility-pattern mining receives increasing attention in the field of data mining recently. With the flourishing development of utility-pattern mining, most studies focused on the efficiency problem by considering the efficient data structure to compress the original data and pruning strategies to reduce the search space for knowledge discovery. However, those approaches can only handle the binary situation, thus the discovered knowledge cannot be represented as the linguistic variables. Previous works have addressed this problem by introducing the generic approaches to find the high fuzzy utility itemsets in a small database. In real-world situations, the dataset may be very large, and it is costly to mine all the required information from a very large database. In this paper, we first present a HFUI-GA framework to discover the high fuzzy utility itemsets in a limited time. Several improvement strategies are also proposed to speed up the evolutionary progress. Experiments are then conducted to show the performance of the variants of the designed HFUI-GA framework in terms of number of the discovered high fuzzy utility itemsets (HFUIs) and the results are convincing to show that the designed GA-based HFUI-GA framework is a promising solution to mine for HFUIs.

In the field of data mining, many utility-oriented mining approaches have been extensively studied. Previous studies have, however, the limitation that they rarely consider the inherent correlation of items among the discovered patterns. For example, from the purchase behavior, a high-utility group of products (w.r.t. multi-products) may contain the items with both high or low utility. This pattern is also considered as a valuable pattern even if they may not be highly correlated, or even happened together by the chance. In this paper, we propose an efficient utility mining approach namely non-redundant Correlated high-Utility Pattern Miner (CoUPM) by considering both strong positive correlation and profitable value of the products. The derived patterns with high utility and strong correlation can lead to more insightful availability than those patterns only have high utility values. The utility-list structure is maintained and applied to store necessary information of correlation and utility. Several pruning strategies are further developed to improve the efficiency for discovering the desired patterns. Experimental results show that the non-redundant correlated high-utility patterns have more effectiveness than some other kinds of patterns. Moreover, the proposed CoUPM algorithm significantly outperforms the state-of-the-art algorithm.

Mining erasable itemsets is a problem derived from the production planning of the manufacturing industry. In the past, the erasable-itemset mining with product insertion has been designed. In this paper, we further consider the maintenance problem from product deletion. We propose an efficient method to solve it. The method is based on the concept of pre-large itemsets to maintain the correct results for product deletion. It improves the efficiency of the mining process by further reducing the number of times required for rescanning the database. When the ratio of the number of deleted products over the total number of products in the original database is less than a certain degree, there will be no need to rescan the original product database for maintaining the correct mining results. Finally, the experiments are made to evaluate the performance of the proposed approach.

Erasable-itemset mining used in production planning identifies itemsets (or components) that, if removed, would not affect profits. Formally, an itemset is erasable if its gain ratio is equal to or smaller than a given maximum gain-ratio threshold r. Since new products with different components may be added, the original batch algorithm will waste time in gathering up-to-date erasable itemsets. In this paper, we propose the concept of the ε-quasi-erasable itemsets and use it to improve mining performance. The itemsets in both the original database and the new product can then be divided into erasable, ε-quasi-erasable, and nonerasable. Thus, there are nine combinations that are then processed in different ways. Experiments are finally made to verify the performance.

In recent years, mining high-utility itemsets (HUIs) has become as a key topic in data mining. However, most of the developed algorithms assume the unrealistic situations that unit profits of items remain unchanged over time. But in real-life situations, the profit of an item or itemset varies as a function of cost prices, sales prices and sales strategies. In this paper, a novel framework for mining HUIs with two algorithms under various Discount strategies (HUID) are introduced. HUID-tp is based on various discount strategies and a novel downward closure property to mine the complete set of HUIs. HUID-Miner is an algorithm relying on a compact data structure (Positive-and-Negative Utility-list, PNU-list) and new pruning strategies to efficiently discover HUIs without candidate generation, while considerably reducing the size of the search space. Furthermore, a strategy named Estimated Utility Co-occurrence Strategy which stores the relationships between 2-itemsets is also adopted in the proposed improvement HUID-EMiner algorithm to speed up computation. An extensive experimental study carried on several real-life datasets shows the performance of the proposed algorithms.

Efficient shortest path calculation has been studied extensively, in particular, in the distributed environment. However, preserving privacy in the cloud environment has just attracted latest attention. To preserve fixed-pattern one-neighborhood privacy in the cloud, current approach requires the calculation of all-pairs shortest paths in advance, which is time consuming for large graphs. In addition, specific paths that are sensitive and require hiding the source and destination vertices are not well addressed. In this work, we propose a new flexible k-neighborhood privacy-protection and efficient shortest distance computation scheme for sensitive shortest paths in the cloud environment. Combining the construction of k-skip shortest path sub-graphs, sensitive vertex adjustment, vertex hierarchy labeling and bottom-up partitioning techniques, the proposed approach not only subsumes one-neighborhood privacy but also provides efficient partitioning and query processing for sensitive shortest paths. Numerical experiments demonstrating the characteristics of proposed approach are presented.

Location and local service is one of the hottest bunches of applications in recent years, due to the proliferation of Global Position System (GPS) and mobile web search technology. Spatial queries retrieving neighboring Point-Of-Interests (POI) require actual user locations for services. However, exposing the physical location of querier to service system may pose privacy threat to users, if malicious adversary has access to the system. To hinder the service system from obtaining the "true" location of querier, current obfuscation-based approach requires a trusted third party anonymizer. As for the data-encryption-based and cPIR-based approaches, they incur costly computation overheads. Although the secure hardware-aided PIR-based technique has been shown to be superior to formers, it did not consider the characteristics of data distribution of searching domain. To deal with the problem of non-uniform data distribution and efficient retrieval, we propose a scheme, MHBL, based on Hilbert space-filling curve and flexible multi-layer grids for efficient storage and retrieval of POI data, so that improved performance of PIR-based techniques could be achieved. Numerical experiments demonstrate that the proposed technique indeed deliver better efficiency under various criteria.

In this paper, we try to automatically induce the membership functions appropriate for the TS fuzzy model. A GA-based learning algorithm is thus proposed to achieve the purpose. The proposed approach considers the shapes of membership functions in fitness evaluation in addition to the accuracy. The shapes of membership functions are evaluated by the overlap and coverage factors, which are used to avoid the bad types of membership functions. The experimental results show that the proposed approach can derive the membership functions in the Takagi-Sugeno system with low errors and good shapes.

In the past, we proposed a Fast Updated FP-tree (FUFP-tree) structure to efficiently handle new transactions and to make the tree update process become easier. In this paper, we attempt to modify the FUFP-tree construction based on the concept of pre-large itemsets. Pre-large itemsets are defined by a lower support threshold and an upper support threshold. The proposed approach can achieve a good execution time for tree construction especially when each time a small number of transactions are inserted. Experimental results also show that the proposed Pre-FUFP maintenance algorithm has a good performance for incrementally handling new transactions.

In previous work we proposed incremental mining algorithms for maintenance of sequential patterns based on the concept of pre-large sequences as records were inserted or deleted. Although maintenance of sequential patterns for record modification can be performed by using the deletion procedure and then the insertion procedure, double the computation time of a single procedure is needed. In this paper, we attempt to apply the concept of pre-large sequences to maintain sequential patterns as records are modified. The proposed algorithm does not require rescanning original databases until the accumulative number of modified customer sequences exceeds a safety bound derived by a pre-large concept. As databases grow larger, the number of modified customer sequences allowed before database rescanning also needs to grow.

The authors propose a novel mining approach based on the genetic process and an evaluation mechanism to automatically construct an effective fuzzy rule base. The proposed approach consists of three phases: fuzzy-rule generating, fuzzy-rule encoding and fuzzy-rule evolution. In the fuzzy-rule generating phase, a number of fuzzy rules are randomly generated. In the fuzzy-rule encoding phase, all the rules generated are translated into fixed-length bit strings to form an initial population. In the fuzzy-rule evolution phase, genetic operations and credit assignment are applied at the rule level. The proposed mining approach chooses good individuals in the population for mating, gradually creating better offspring fuzzy rules. A concise and compact fuzzy rule base is thus constructed effectively without human expert intervention.

We previously proposed an incremental mining algorithm for maintenance of sequential patterns based on the concept of pre-large sequences as new records were inserted. In this paper we attempt to apply the concept of pre-large sequences to maintain sequential patterns as records are deleted. Pre-large sequences are defined by a lower support threshold and an upper support threshold. They act as buffers to avoid the movements of sequential patterns directly from large to small and, vice-versa. Our proposed algorithm does not require rescanning original databases until the accumulative amount of deleted customer sequences exceeds a safety bound, which depends on database size. As databases grow larger, the number of deleted customer sequences allowed before database rescanning is required also grows. The proposed approach is thus efficient for a large database.

We have modified the original version space strategy in order to learn disjunctive concepts incrementally and without saving past training instances. The algorithm time complexity is also analyzed, and its correctness is proven.

Article Free Access Share on Learning fuzzy knowledge from training examples Authors: Tzung-Pei Hong Department of Information Management, I-Show University, kaohsiung, 84008, Taiwan, R.O.C Department of Information Management, I-Show University, kaohsiung, 84008, Taiwan, R.O.CView Profile , Chai-Ying Lee Industrial Technology Research Institute, Chutung, hsinchu, Taiwan 310, R.O.C Industrial Technology Research Institute, Chutung, hsinchu, Taiwan 310, R.O.CView Profile Authors Info & Claims CIKM '98: Proceedings of the seventh international conference on Information and knowledge managementNovember 1998Pages 161–166https://doi.org/10.1145/288627.288653Published:01 November 1998Publication History 7citation751DownloadsMetricsTotal Citations7Total Downloads751Last 12 Months33Last 6 weeks5 Get Citation AlertsNew Citation Alert added!This alert has been successfully added and will be sent to:You will be notified whenever a record that you have chosen has been cited.To manage your alert preferences, click on the button below.Manage my AlertsNew Citation Alert!Please log in to your account Save to BinderSave to BinderCreate a New BinderNameCancelCreateExport CitationPublisher SiteeReaderPDF

This paper generalizes the learning strategy of version space to manage noisy and uncertain training data. A new learning algorithm is proposed that consists of two main phases: searching and pruning. The searching phase generates and collects possible candidates into a large set; the pruning then prunes this set according to various criteria to find a maximally consistent version space. When the training instances cannot completely be classified, the proposed learning algorithm can make a trade-off between including positive training instances and excluding negative ones according to the requirements of different application domains. Furthermore, suitable pruning parameters are chosen according to a given time limit, so the algorithm can also make a trade-off between time complexity and accuracy. The proposed learning algorithm is then a flexible and efficient induction method that makes the version space learning strategy more practical.

Applies the technique of parallel processing to concept learning. A parallel version-space learning algorithm based upon the principle of divide-and-conquer is proposed. Its time complexity is analyzed to be O(k log/sub 2/n) with n processors, where n is the number of given training instances and k is a coefficient depending on the application domains. For a bounded number of processors in real situations, a modified parallel learning algorithm is then proposed. Experimental results are then performed on a real learning problem, showing that our parallel learning algorithm works, and being quite consistent with the results of theoretical analysis. We conclude that when the number of training instances is large, it is worth learning in parallel because of its faster execution.>