Jerry Chun-Wei Lin

EAAI Journal 2026 Journal Article

Damped window based high occupancy pattern mining with one scanning of data streams

• Myungha Cho
• Hanju Kim
• Hyeonmo Kim
• Taewoong Ryu
• Chanhee Lee
• Heonho Kim
• Bay Vo
• Jerry Chun-Wei Lin

Occupancy-based pattern mining has been researched and developed by researchers as an interesting field of data mining. It discovers high occupancy patterns that have a high proportion in their transactions. The results of the occupancy-based approach are useful for deriving hidden knowledge because the patterns are more valuable compared to traditional frequent patterns. Recently, an incremental method for high occupancy patterns has been suggested to handle transactions inserted in real-time. However, it often generates results that are not realistic. Even if it is suitable for processing dynamic databases, it is not appropriate to make intelligent decisions because the values of old and new transactions are deemed identical. In this paper, we propose an efficient approach for mining high occupancy patterns, considering the flow of time, with the damped window model. By assigning greater importance to recent data, our method allows for the discovery of more meaningful results following the latest trends. The patterns generated by the novel approach encourage users to intelligently discover occupancy-driven patterns over time-sensitive databases. In addition, the designed method relies on a list-based data structure that increases efficiency without generating candidates, and it utilizes a new pruning strategy to greatly reduce the search space. Extensive experiments using both real and synthetic datasets demonstrate that the proposed approach shows better performance in accordance with varying thresholds and exhibits high efficiency consistently, regardless of the increase in the number of transactions and discrete items.

AAAI Conference 2026 System Paper

SmartEyes: Plug-and-Play Event Detection for Retail Loss Prevention

• Pi-Wei Chen
• Jerry Chun-Wei Lin
• Barış Fahri Kahrıman
• Zih-Ching Chen
• Rafał Cupek
• Marek Drewniak

Event detection is essential for surveillance, particularly in retail loss prevention, where accurate and timely monitoring is critical. Vision Language Models (VLMs) provide strong generalization but are inefficient at processing full video streams and are prone to hallucinations induced by redundant frames. We present SmartEyes, a plug-and-play system for real-time retail surveillance. SmartEyes introduces the Perception Cognition Focusing (PCF) framework, which combines lightweight perception with semantic triggering to isolate two keyframes (customer contact and departure) and constrains the VLMs to a focused differencing task. This design reduces hallucination by 44% compared to vanilla VLMs. From the demonstrated retail application, the proposed perception-to-reasoning pipeline is general and directly extends to industrial environments that require reliable event detection and real-time decision-making. Our demo includes a user-friendly Region of Interest (ROI) selection interface and live CCTV monitoring, producing accurate alerts within 1–2 seconds on a single RTX 4080 GPU. This lightweight framework design enables efficient deployment to broader industrial applications.

JBHI Journal 2026 Journal Article

Towards Early Diagnosis and Intervention: An Ensemble Voting Model for Precise Vital Sign Prediction in Respiratory Disease

• Usman Ahmed
• Jerry Chun-Wei Lin
• Gautam Srivastava

Worldwide, cardiovascular and chronic respiratory diseases have approximately million deaths each year. Evidence indicates that the ongoing COVID-19 pandemic directly contributed to increase blood pressure, cholesterol, and blood glucose levels. Timely screening of critical physiological vital signs benefits both healthcare providers and individuals by detecting potential health issues. This study aimed to implement a machine learning-based prediction and classification system to forecast vital signs associated with cardiovascular and chronic respiratory diseases. The system predicts patients' health status and notifies medical professionals when necessary. Utilizing real-world data, a linear regression model inspired by the Facebook Prophet model was utilized to predict vital signs for the next 180 seconds. With lead time of 180 seconds, medical professionals can potentially save patients' lives through early diagnosis of their health conditions. For this purpose, Naïve Bayes classification model, Support Vector Machine model, a Random Forest model, and genetic programming-based hyper tunning were employed. The proposed model outperforms previous attempts for vital sign prediction. Compared with alternative methods, the Facebook Prophet model had the mean sqaure error for predicting vital signs. Hyperparameter tunning was utilized to refine model, yielding improved short- and long-term outcomes for each vital sign. From the results, it showed that the designed model acheives higher F-measure performance. The incorporation of additional elements, such as momentum indicators, can increase the flexibility of the model with calibration. The findings of this study demonstrated that the proposed model is more accurate in predicting vital signs and trends.

ECAI Conference 2025 Conference Paper

BakuFlow: A Streamlining Semi-Automatic Label Generation Tool

• Jerry Chun-Wei Lin
• Pi-Wei Chen
• Rafal Cupek

Accurately labeling (or annotation) data is still a bottleneck in computer vision, especially for large-scale tasks where manual labeling is time-consuming and error-prone. While tools like LabelImg can handle the labeling task, some of them still require annotators to manually label each image. In this paper, we introduce BakuFlow, a streamlining semi-automatic label generation tool. Key features include (1) a live adjustable magnifier for pixel-precise manual corrections, improving user experience; (2) an interactive data augmentation module to diversify training datasets; (3) label propagation for rapidly copying labeled objects between consecutive frames, greatly accelerating annotation of video data; and (4) an automatic labeling module powered by a modified YOLOE framework. Unlike the original YOLOE, our extension supports adding new object classes and any number of visual prompts per class during annotation, enabling flexible and scalable labeling for dynamic, real-world datasets. These innovations make BakuFlow especially effective for object detection and tracking, substantially reducing labeling workload and improving efficiency in practical computer vision and industrial scenarios.

JBHI Journal 2025 Journal Article

Guest Editorial: Current Trends and Future Directions in Biomedical Data Science

• Hossein Moosaei
• Roohallah Alizadehsani
• Mario Rosario Guarracino
• Jerry Chun-Wei Lin

JBHI Journal 2025 Journal Article

Label-Aware Dual Graph Neural Networks for Multi-Label Fundus Image Classification

• Yanbei Liu
• Xinwen Peng
• Xin Wei
• Lei Geng
• Fang Zhang
• Zhitao Xiao
• Jerry Chun-Wei Lin

Fundus disease is a complex and universal disease involving a variety of pathologies. Its early diagnosis using fundus images can effectively prevent further diseases and provide targeted treatment plans for patients. Recent deep learning models for classification of this disease are gradually emerging as a critical research field, which is attracting widespread attention. However, in practice, most of the existing methods only focus on local visual cues of a single image, and ignore the underlying explicit interaction similarity between subjects and correlation information among pathologies in fundus diseases. In this paper, we propose a novel label-aware dual graph neural networks for multi-label fundus image classification that consists of population-based graph representation learning and pathology-based graph representation learning modules. Specifically, we first construct a population-based graph by integrating image features and non-image information to learn patient's representations by incorporating associations between subjects. Then, we represent pathologies as a sparse graph where its nodes are associated with pathology-based feature vectors and the edges correspond to probability of the co-occurrence of labels to generate a set of classifier scores by the propagation of multi-layer graph information. Finally, our model can adaptively recalibrate multi-label outputs. Detailed experiments and analysis of our results show the effectiveness of our method compared with state-of-the-art multi-label fundus image classification methods.

ECAI Conference 2025 Conference Paper

Visualizing Clickstream Prediction Logic and Utility

• Danny Yu-Chung Wang
• Lars Arne Jordanger
• Jerry Chun-Wei Lin

Understanding user behavior patterns from clickstream data is critical for improving e-commerce decision-making. In this study, we propose a novel visualization framework called SankeyX that connects user interaction sequences, model predictions, SHAP-based feature attributions, and business utility into a unified Sankey-style diagram. It enables users to trace how behavioral patterns contribute to model outcomes and assess their financial impact by a utility matrix. In the case study, we also demonstrate its ability to reveal dominant purchasing patterns by using a real world dataset. This method bridges the gap between explainable AI and decision-making for clickstream prediction.

JBHI Journal 2024 Journal Article

A Scalable Federated Learning Approach for Collaborative Smart Healthcare Systems With Intermittent Clients Using Medical Imaging

• Farhan Ullah
• Gautam Srivastava
• Heng Xiao
• Shamsher Ullah
• Jerry Chun-Wei Lin
• Yue Zhao

The healthcare industry is one of the most vulnerable to cybercrime and privacy violations because health data is very sensitive and spread out in many places. Recent confidentiality trends and a rising number of infringements in different sectors make it crucial to implement new methods that protect data privacy while maintaining accuracy and sustainability. Moreover, the intermittent nature of remote clients with imbalanced datasets poses a significant obstacle for decentralized healthcare systems. Federated learning (FL) is a decentralized and privacy-protecting approach to deep learning and machine learning models. In this article, we implement a scalable FL framework for interactive smart healthcare systems with intermittent clients using chest X-ray images. Remote hospitals may have imbalanced datasets with intermittent clients communicating with the FL global server. The data augmentation method is used to balance datasets for local model training. In practice, some clients may leave the training process while others join due to technical or connectivity issues. The proposed method is tested with five to eighteen clients and different testing data sizes to evaluate performance in various situations. The experiments show that the proposed FL approach produces competitive results when dealing with two distinct problems, such as intermittent clients and imbalanced data. These findings would encourage medical institutions to collaborate and use rich private data to quickly develop a powerful patient diagnostic model.

JBHI Journal 2024 Journal Article

Graph Attention-Based Curriculum Learning for Mental Healthcare Classification

• Usman Ahmed
• Jerry Chun-Wei Lin
• Gautam Srivastava

Current research has examined the use of user-generated data from online media to identify and diagnose depression as a serious mental health issue that can significantly impact an individual's daily life. To this end, many studies examined words in personal statements to identify depression. In addition to aiding in the diagnosis and treatment of depression, this study uses and utilizes a Graph Attention Network (GAT) model for the classification of depression from online media. The model is based on masked self-attention layers, that assigns different weight to each node in a neighborhood without costly matrix operations. In addition, an emotion lexicon was extended using hypernyms to improve the model performance. Furthermore, embedding of the model was used to illustrate the contribution of the activated words to each symptom and to obtain qualitative agreement from psychiatrists. This technique uses previously learned embedding to illustrate the contribution of activated words to depressive symptoms in online forums. A significant improvement was observed in the model's performance through the use of the lexicon extension method, resulting in an increase in the ROC performance. The performance was also enhanced by an increase in vocabulary and the adoption of a graph-based curriculum. The lexicon expansion method involves the generation of additional words with similar semantic attributes, utilizing similarity metrics to reinforce lexical features. The graph-based curriculum learning also utilized to handle more challenging training samples, allowing the model to develop increasing expertise in learning complex correlations between input data and output labels.

JBHI Journal 2023 Journal Article

Hyper-Graph Attention Based Federated Learning Methods for Use in Mental Health Detection

• Usman Ahmed
• Jerry Chun-Wei Lin
• Gautam Srivastava Srivastava

Internet-Delivered Psychological Treatment (IDPT) has become necessary in the medical field. Deep neural networks (DNNs) require large, diverse patient populations to train models that achieve clinician-level performance. However, DNN models trained on limited datasets have poor clinical performance when used in a new location with different data. Thus, increasing the availability of diverse as well as distinct training data is vital. This study proposes a structural hypergraph as well as an emotional lexicon for word representation. An embedding model based on federated learning was developed for mental health symptom detection. The model treats text data as a collection of consecutive words. The model then learns a low-dimensional continuous vector while maintaining contextual linkage. The generated models with attention-based mechanisms as well as federated learning are then tested experimentally. Our strategy is suitable for vocabulary diversification, grammatical word representation, as well as dynamic lexicon analysis. The goal is to create semantic word representations using an attention network model. Later, clinical processes are used to mark the text by embedding it. Experimental results show the encoding of emotional words using the structural hypergraph. The 0. 86 ROC was achieved using the bidirectional LSTM architecture with an attention mechanism.

JBHI Journal 2023 Journal Article

Multi-Aspect Deep Active Attention Network for Healthcare Explainable Adoption

• Usman Ahmed
• Jerry Chun-Wei Lin
• Gautam Srivastava

Depression is a serious illness that significantly affects the lives of those affected. Recent studies have looked at the possibility of detecting and diagnosing this mental disorder using user-generated data from various forms of online media. Therefore, we address the issue of detecting sadness in social media by focusing on terms in personal remarks. To overcome the limitations in classifying depression texts, this study aims to develop attention networks that use covert levels of self-attention. Since nodes/words can express properties/emotions of their neighbors, this paper naturally assigns each node in a neighborhood a weight without performing costly matrix operations such as similarity or network architecture knowledge. This paper extends the emotion lexicon by using hypernyms. For this reason, our method is superior to the performance of other designs. According to the results of our experiments, the emotion lexicon combined with an attention network achieves an ROC of 0. 87 while maintaining its interpretability and transparency level. Subsequently, the learned embedding is used to display the contribution of each symptom to the activated word, and the psychiatrist is polled to obtain his qualitative agreement with this representation. By using unlabeled forum language, the method increases the rate at which depression symptoms can be identified from information in Internet forums.

EAAI Journal 2023 Journal Article

Prediction and control of water quality in Recirculating Aquaculture System based on hybrid neural network

• Junchao Yang
• Lulu Jia
• Zhiwei Guo
• Yu Shen
• Xianwei Li
• Zhenping Mou
• Keping Yu
• Jerry Chun-Wei Lin

In the Recirculating Aquaculture Systems (RAS), the control of water quality indices remains essential to survival and growth of aquaculture objects. This requires effect prediction of future water status in advance, which can be adopted to help the generation of following control strategies. However, conventional methods of water quality prediction were mostly dependent on redundant parameters of model, which leads to inefficiency and low accuracy. In addition, the complexity of the RAS multi-units requires intelligent control of the water quality unit. Thus, a prediction and control framework for predicting water quality in RAS is proposed in this paper. Specifically, a hybrid deep learning structure which combines the Convolutional Neural Network (CNN), Gated Recurrent Unit (GRU) and Attention mechanism is presented. To begin with, the CNN is utilized to extract local features for different timestamped water quality parameter. After the local features have been extracted, the proposed GRU model replicates the global sequential features of the parameters. The attention mechanism is then applied to focus on more critical features to promote the efficiency and accuracy of prediction. Finally, to demonstrate the efficiency and stability of the prediction and control framework with the mixture of CNN, GRU and Attention (PC-CGA), multiple groups of experiments and evaluations are carried out in a medium size RAS.

JBHI Journal 2022 Journal Article

Secure Collaborative Augmented Reality Framework for Biomedical Informatics

• Youcef Djenouri
• Asma Belhadi
• Gautam Srivastava
• Jerry Chun-Wei Lin

Augmented reality is currently of interest in biomedical health informatics. At the same time, several challenges have appeared, in particular with the rapid progress of smart sensor technologies, and medical artificial intelligence. This yields the necessity of new needs in biomedical health informatics. Collaborative learning and privacy are just some of the challenges of augmented reality technology in biomedical health informatics. This paper introduces a novel secure collaborative augmented reality framework for biomedical health informatics-based applications. Distributed deep learning is performed across a multi-agent system platform. The privacy strategy is then developed for ensuring better communications of the different intelligent agents in the system. In this research work, a system of multiple agents is created for the simulation of the collective behaviours of the smart components of biomedical health informatics. Augmented reality is also incorporated for better visualization of medical patterns. A novel privacy strategy based on blockchain is investigated for ensuring the confidentiality of the learning process. Experiments are conducted on real use cases of the biomedical segmentation process. Our strong experimental analysis reveals the strength of the proposed framework when directly compared to state-of-the-art biomedical health informatics solutions.

TIST Journal 2021 Journal Article

RHUPS

• Yoonji Baek
• Unil Yun
• Heonho Kim
• Hyoju Nam
• Hyunsoo Kim
• Jerry Chun-Wei Lin
• Bay Vo
• Witold Pedrycz

Databases that deal with the real world have various characteristics. New data is continuously inserted over time without limiting the length of the database, and a variety of information about the items constituting the database is contained. Recently generated data has a greater influence than the previously generated data. These are called the time-sensitive non-binary stream databases, and they include databases such as web-server click data, market sales data, data from sensor networks, and network traffic measurement. Many high utility pattern mining and stream pattern mining methods have been proposed so far. However, they have a limitation that they are not suitable to analyze these databases, because they find valid patterns by analyzing a database with only some of the features described above. Therefore, knowledge-based software about how to find meaningful information efficiently by analyzing databases with these characteristics is required. In this article, we propose an intelligent information system that calculates the influence of the insertion time of each batch in a large-scale stream database by applying the sliding window model and mines recent high utility patterns without generating candidate patterns. In addition, a novel list-based data structure is suggested for a fast and efficient management of the time-sensitive stream databases. Moreover, our technique is compared with state-of-the-art algorithms through various experiments using real datasets and synthetic datasets. The experimental results show that our approach outperforms the previously proposed methods in terms of runtime, memory usage, and scalability.

EAAI Journal 2020 Journal Article

Erasable pattern mining based on tree structures with damped window over data streams

• Yoonji Baek
• Unil Yun
• Heonho Kim
• Hyoju Nam
• Gangin Lee
• Eunchul Yoon
• Bay Vo
• Jerry Chun-Wei Lin

Several pattern mining methods have been proposed to process dynamic data streams because the data generated in industrial fields is continually accumulated. Erasable pattern mining techniques for processing dynamic data streams are needed to discover erasable patterns from dynamic data streams. In previous erasable pattern mining approaches suggested for dynamic data streams, all data are considered to have the same importance regardless of its timestamp. However, dynamic data streams have the characteristic that the new data is relatively more significant than the old data. In erasable pattern mining, one of the desired techniques is an approach in consideration of such characteristic of data streams. For this reason, we propose an erasable pattern mining algorithm over dynamic data streams based on the damped window model. Since the suggested technique considers the new data more important than the previous data, it can find more useful erasable patterns. In addition, erasable pattern mining based on the damped window model is conducted efficiently by employing the tree and table structures. In performance test, we present that our pruning techniques remove unnecessary operations related to invalid erasable patterns efficiently from damped-window-based data streams. Performance evaluation results using real datasets and synthetic datasets show that the proposed approach has good performance with regard to as execution time, pattern generation, and scalability by comparing between the suggested technique and the state of the art algorithms.

EAAI Journal 2020 Journal Article

Space–time series clustering: Algorithms, taxonomy, and case study on urban smart cities

• Asma Belhadi
• Youcef Djenouri
• Kjetil Nørvåg
• Heri Ramampiaro
• Florent Masseglia
• Jerry Chun-Wei Lin

This paper provides a short overview of space–time series clustering, which can be generally grouped into three main categories such as: hierarchical, partitioning-based, and overlapping clustering. The first hierarchical category is to identify hierarchies in space–time series data. The second partitioning-based category focuses on determining disjoint partitions among the space–time series data, whereas the third overlapping category explores fuzzy logic to determine the different correlations between the space–time series clusters. We also further describe solutions for each category in this paper. Furthermore, we show the applications of these solutions in an urban traffic data captured on two urban smart cities (e. g. , Odense in Denmark and Beijing in China). The perspectives on open questions and research challenges are also mentioned and discussed that allow to obtain a better understanding of the intuition, limitations, and benefits for the various space–time series clustering methods. This work can thus provide the guidances to practitioners for selecting the most suitable methods for their used cases, domains, and applications.

EAAI Journal 2019 Journal Article

A Bi-LSTM mention hypergraph model with encoding schema for mention extraction

• Jerry Chun-Wei Lin
• Yinan Shao
• Yujie Zhou
• Matin Pirouz
• Hsing-Chung Chen

Natural language processing is a technique to process data such as text and speech. Some fundamental research includes named-entity recognition, which recognizes name entities (i. e. , persons, companies) from texts; semantic parsing, which is used to convert a natural language utterance to the representation of logical form; and co-reference resolution, which extracts nouns (including pronouns, noun phrases) pointing to the same reference body. In this paper, we mainly focus on the task of mention extraction, which extract and classify overlapping or nested structure mentions. We proposed a neural-encoded mention-hypergraph (NEMH) model to use hypergraph to model overlapping or nested structure mentions and use neural networks to extract features for hypergraph automatically. Unlike the existing approaches, our hypergraph model can effectively capture nested mention entities with unlimited lengths. Also, the proposed model is highly scalable and the time complexity of the proposed model is linear in the number of mention classes and the number of input words. Extensive experiments are conducted on several standard datasets to demonstrate the effectiveness of the proposed model.

EAAI Journal 2019 Journal Article

Mining of skyline patterns by considering both frequent and utility constraints

• Jerry Chun-Wei Lin
• Lu Yang
• Philippe Fournier-Viger
• Tzung-Pei Hong

Association-rule mining (ARM) or frequent itemset mining (FIM) is the most fundamental task in knowledge discovery, which is used to find the occurrence frequency of the item/sets in transactional database. The other factors such as weight, interestingness or unit profit of the items are not considered whether in ARM or FIM. To reveal more information, high-utility itemset mining (HUIM) was designed to consider both quantity and unit profit of items to discover the high-utility itemsets (HUIs). Several algorithms for FIM or HUIM were extensively studied but fewer works concern both frequency and utility together to provide better solutions in decision-making. In the past, the SKYMINE algorithm was designed to find the skyline frequent-utility patterns (SFUPs). A SFUP is a non-dominated pattern, in which each solution dominates the others by considering the aspects of frequency and utility. The SKYMINE algorithm needs, however, amounts of computation to level-wisely discover the SFUPs. In this paper, an efficient utility-list structure is used instead of the UP-tree structure used in SKYMINE to mine the SFUPs. Two algorithms are respectively designed by using the depth-first search (called SKYFUP-D) and breath-first search (SKYFUP-B) to mine the SFUPs. An efficient structure is also designed to record the maximal utility of the potential itemsets, thus reducing the computations for finding the SFUPs in the search space. Extensive experiments are conducted on several real-world and simulated datasets and the results indicate that the designed two algorithms have better performance than that of the state-of-the-art SKYMINE algorithm in terms of runtime, memory usage, search space size and the scalability.

EAAI Journal 2018 Journal Article

Efficiently updating the discovered high average-utility itemsets with transaction insertion

• Jerry Chun-Wei Lin
• Shifeng Ren
• Philippe Fournier-Viger
• Jeng-Shyan Pan
• Tzung-Pei Hong

High-utility itemset mining (HUIM) is an extension of frequent-itemset mining (FIM) but considers the unit profit and quantity of items to discover the set of high-utility itemsets (HUIs). Traditionally, the utility of an itemset is the summation of the utilities of the itemset in all the transactions regardless of its length. This approach is, however, inappropriate in real-world applications since the utility of the itemset increases along with the number of items within it. High average-utility itemset mining (HAUIM) was designed to provide more reasonable utility measure by taking the size of the itemset into account. Existing algorithms can only handle, however, the static database and unsuitable for the dynamic environment since the size of data is frequently changed in real-life situations. In this paper, an incremental high-average utility pattern mining (IHAUPM) algorithm is presented to handle the incremental database with transaction insertion. The well-known fast updated (FUP) concept in the FIM is modified to adopt the designed algorithm, thus efficiently updating the discovered HAUIs. Based on the designed model for HAUIM with transaction insertion, the proposed IHAUPM algorithm can easily only handle the inserted transactions. Experiments are carried on six datasets and the results showed that the designed algorithm has better performance than the state-of-the-art algorithms performing in the batch manner.

EAAI Journal 2017 Journal Article

Extracting recent weighted-based patterns from uncertain temporal databases

• Wensheng Gan
• Jerry Chun-Wei Lin
• Philippe Fournier-Viger
• Han-Chieh Chao
• Jimmy Ming-Tai Wu
• Justin Zhan

Weighted Frequent Itemset Mining (WFIM) has been proposed as an extension of frequent itemset mining that considers not only the frequency of items but also their relative importance. However, using WFIM algorithms in real applications raises some problems. First, they do not consider how recent the patterns are. Second, traditional WFIM algorithms cannot handle uncertain data, although this type of data is common in real-life. To address these limitations, this paper introduces the concept of Recent High Expected Weighted Itemset (RHEWI), which considers the recency, weight and uncertainty of patterns. By considering these three factors, more up-to-date and relevant results are found. A projection-based algorithm named RHEWI-P is presented to mine RHEWIs using a novel upper-bound downward closure (UBDC) property. An improved version of this algorithm called RHEWI-PS is further proposed based on a novel sorted upper-bound downward closure (SUBDC) property for pruning unpromising candidate itemsets early. An experimental evaluation against the state-of-the-art HEWI-Uapriori algorithm was carried out on both real-world and synthetic datasets. Results show that the proposed algorithms are highly efficient and are acceptable for mining the desired patterns.

EAAI Journal 2017 Journal Article

Mining of frequent patterns with multiple minimum supports

• Wensheng Gan
• Jerry Chun-Wei Lin
• Philippe Fournier-Viger
• Han-Chieh Chao
• Justin Zhan

Frequent pattern mining (FPM) is an important topic in data mining for discovering the implicit but useful information. Many algorithms have been proposed for this task but most of them suffer from an important limitation, which relies on a single uniform minimum support threshold as the sole criterion to identify frequent patterns (FPs). Using a single threshold value to assess the usefulness of all items in a database is inadequate and unfair in real-life applications since each item is different and not all items should be treated as the same. Several algorithms have been developed for mining FPs with multiple minimum supports but most of them suffer from the time-consuming problem and require a large amount of memory. In this paper, we address this issue by introducing the novel approach named F requent P attern mining with M ultiple minimum supports from the E numeration-tree (FP-ME). In the developed Set- E numeration -tree with M ultiple minimum supports (ME-tree) structure, a new sorted downward closure (SDC) property of FPs and the least minimum support (LMS) concept with multiple minimum supports are used to effectively prune the search space. The proposed FP-ME algorithm can directly discover FPs from the ME-tree without candidate generation. Moreover, an improved algorithm, named FP-MEDiffSet, is also developed based on the DiffSet concept, to further increase mining performance. Substantial experiments on both real-life and synthetic datasets show that the proposed algorithms can not only avoid the “rare item problem”, but also efficiently and effectively discover the complete set of FPs in transactional databases while considering multiple minimum supports and outperform the state-of-the-art CFP-growth++ algorithm in terms of execution time, memory usage and scalability.

EAAI Journal 2016 Journal Article

A sanitization approach for hiding sensitive itemsets based on particle swarm optimization

• Jerry Chun-Wei Lin
• Qiankun Liu
• Philippe Fournier-Viger
• Tzung-Pei Hong
• Miroslav Voznak
• Justin Zhan

Privacy-preserving data mining (PPDM) has become an important research field in recent years, as approaches for PPDM can discover important information in databases, while ensuring that sensitive information is not revealed. Several algorithms have been proposed to hide sensitive information in databases. They apply addition and deletion operations to perturb an original database and hide the sensitive information. Finding an appropriate set of transactions/itemsets to be perturbed for hiding sensitive information while preserving other important information is a NP-hard problem. In the past, genetic algorithm (GA)-based approaches were developed to hide sensitive itemsets in an original database through transaction deletion. In this paper, a particle swarm optimization (PSO)-based algorithm called PSO2DT is developed to hide sensitive itemsets while minimizing the side effects of the sanitization process. Each particle in the designed PSO2DT algorithm represents a set of transactions to be deleted. Particles are evaluated using a fitness function that is designed to minimize the side effects of sanitization. The proposed algorithm can also determine the maximum number of transactions to be deleted for efficiently hiding sensitive itemsets, unlike the state-of-the-art GA-based approaches. Besides, an important strength of the proposed approach is that few parameters need to be set, and it can still find better solutions to the sanitization problem than GA-based approaches. Furthermore, the pre-large concept is also adopted in the designed algorithm to speed up the evolution process. Substantial experiments on both real-world and synthetic datasets show that the proposed PSO2DT algorithm performs better than the Greedy algorithm and GA-based algorithms in terms of runtime, fail to be hidden (F-T-H), not to be hidden (N-T-H), and database similarity (DS).

EAAI Journal 2016 Journal Article

Fast algorithms for hiding sensitive high-utility itemsets in privacy-preserving utility mining

• Jerry Chun-Wei Lin
• Tsu-Yang Wu
• Philippe Fournier-Viger
• Guo Lin
• Justin Zhan
• Miroslav Voznak

High-Utility Itemset Mining (HUIM) is an extension of frequent itemset mining, which discovers itemsets yielding a high profit in transaction databases (HUIs). In recent years, a major issue that has arisen is that data publicly published or shared by organizations may lead to privacy threats since sensitive or confidential information may be uncovered by data mining techniques. To address this issue, techniques for privacy-preserving data mining (PPDM) have been proposed. Recently, privacy-preserving utility mining (PPUM) has become an important topic in PPDM. PPUM is the process of hiding sensitive HUIs (SHUIs) appearing in a database, such that the resulting sanitized database will not reveal these itemsets. In the past, the HHUIF and MSICF algorithms were proposed to hide SHUIs, and are the state-of-the-art approaches for PPUM. In this paper, two novel algorithms, namely Maximum Sensitive Utility-MAximum item Utility (MSU-MAU) and Maximum Sensitive Utility-MInimum item Utility (MSU-MIU), are respectively proposed to minimize the side effects of the sanitization process for hiding SHUIs. The proposed algorithms are designed to efficiently delete SHUIs or decrease their utilities using the concepts of maximum and minimum utility. A projection mechanism is also adopted in the two designed algorithms to speed up the sanitization process. Besides, since the evaluation criteria proposed for PPDM are insufficient and inappropriate for evaluating the sanitization performed by PPUM algorithms, this paper introduces three similarity measures to respectively assess the database structure, database utility and item utility of a sanitized database. These criteria are proposed as a new evaluation standard for PPUM.

EAAI Journal 2016 Journal Article

Mining high-utility itemsets based on particle swarm optimization

• Jerry Chun-Wei Lin
• Lu Yang
• Philippe Fournier-Viger
• Jimmy Ming-Thai Wu
• Tzung-Pei Hong
• Leon Shyue-Liang Wang
• Justin Zhan

High-utility itemset mining (HUIM) is a critical issue in recent years since it can be used to reveal the profitable products by considering both the quantity and profit factors instead of frequent itemset mining (FIM) or association-rule mining (ARM). Several algorithms have been presented to mine high-utility itemsets (HUIs) and most of the designed algorithms have to handle the exponential search space for discovering HUIs when the number of distinct items and the size of database are very large. In the past, a heuristic HUPEumu-GRAM algorithm was proposed to mine HUIs based on genetic algorithm (GA). For the evolutionary computation (EC) techniques of particle swarm optimization (PSO), it only requires fewer parameters compared to the GA-based approach. Since the traditional PSO mechanism is used to handle the continuous problem, in this paper, the discrete PSO is adopted to encode the particles as the binary variables. An efficient PSO-based algorithm namely HUIM-BPSOsig is proposed to efficiently find HUIs. It first sets the number of discovered high-transaction-weighted utilization 1-itemsets (1-HTWUIs) as the size of a particle based on transaction-weighted utility (TWU) model, which can greatly reduce the combinational problem in evolution process. The sigmoid function is adopted in the updating process of the particles of the designed HUIM-BPSOsig algorithm. Substantial experiments on real-life datasets show that the proposed algorithm has better results compared to the state-of-the-art GA-based algorithm.

EAAI Journal 2015 Journal Article

RWFIM: Recent weighted-frequent itemsets mining

• Jerry Chun-Wei Lin
• Wensheng Gan
• Philippe Fournier-Viger
• Tzung-Pei Hong

In recent years, weighted frequent itemsets mining (WFIM) has become a critical issue of data mining, which can be used to discover more useful and interesting patterns in real-world applications instead of the traditional frequent itemsets mining. Many algorithms have been developed to find weighted frequent itemsets (WFIs) without time-sensitive consideration. The discovered out-of-date information may, however, be meaningless and useless in decision making. In this paper, a novel framework, namely recent weighted-frequent itemsets mining (RWFIM) is proposed to concern both the weight and time-sensitive constraints. A projected-based RWFIM-P algorithm is first proposed for mining the designed recent weighted-frequent itemsets (RWFIs) with weight and time-sensitive consideration. It uses the projection-and-test mechanism to discover RWFIs in a recursive way. Based on the developed RWFIM-P algorithm, the entire database can be projected and divided into several sub-databases according to the currently processed itemset, thus reducing the computational costs and memory requirements. The second RWFIM-PE algorithm is also proposed to improve the performance of the first RWFIM-P algorithm based on the developed Estimated Weight of 2-itemset Pruning (EW2P) strategy to mine the RWFIs without generating the unpromising candidates, thus avoiding the computations of the projection mechanism compared to the first RWFIM-P algorithm. Experiments are conducted to evaluate the performance of the proposed two algorithms compared to the traditional WFIM in terms of execution time, number of generated RWFIs and scalability under varied two minimum thresholds in several real-world and synthetic datasets.