The Synthetic Minority Oversampling Technique (SMOTE) preprocessing algorithm is considered "de facto" standard in the framework of learning from imbalanced data. This is due to its simplicity in the design of the procedure, as well as its robustness when applied to different type of problems. Since its publication in 2002, SMOTE has proven successful in a variety of applications from several different domains. SMOTE has also inspired several approaches to counter the issue of class imbalance, and has also significantly contributed to new supervised learning paradigms, including multilabel classification, incremental learning, semi-supervised learning, multi-instance learning, among others. It is standard benchmark for learning from imbalanced data. It is also featured in a number of different software packages -- from open source to commercial. In this paper, marking the fifteen year anniversary of SMOTE, we reect on the SMOTE journey, discuss the current state of affairs with SMOTE, its applications, and also identify the next set of challenges to extend SMOTE for Big Data problems.

/pdf/smote-for-learning-from-imbalanced-data-progress-and-3slcu62yo6.pdf

SMOTE for learning from imbalanced data: progress and challenges, marking the 15-year anniversary

In this paper, we address the problem of the identification of text in noisy document images. We are especially focused on segmenting and identifying between handwriting and machine printed text because: 1) Handwriting in a document often indicates corrections, additions, or other supplemental information that should be treated differently from the main content and 2) the segmentation and recognition techniques requested for machine printed and handwritten text are significantly different. A novel aspect of our approach is that we treat noise as a separate class and model noise based on selected features. Trained Fisher classifiers are used to identify machine printed text and handwriting from noise and we further exploit context to refine the classification. A Markov Random Field-based (MRF) approach is used to model the geometrical structure of the printed text, handwriting, and noise to rectify misclassifications. Experimental results show that our approach is robust and can significantly improve page segmentation in noisy document collections.

/pdf/machine-printed-text-and-handwriting-identification-in-noisy-sd3e5xo6r7.pdf

Machine printed text and handwriting identification in noisy document images

https://rgu-repository.worktribe.com/preview/512831/VUTTIPITTAYAMONGKOL%202020%20Neighbourhood%20based.pdf

Neighbourhood-based undersampling approach for handling imbalanced and overlapped data

Characteristics extracted from the training datasets of classification problems have proven to be effective predictors in a number of meta-analyses. Among them, measures of classification complexity can be used to estimate the difficulty in separating the data points into their expected classes. Descriptors of the spatial distribution of the data and estimates of the shape and size of the decision boundary are among the known measures for this characterization. This information can support the formulation of new data-driven pre-processing and pattern recognition techniques, which can in turn be focused on challenges highlighted by such characteristics of the problems. This article surveys and analyzes measures that can be extracted from the training datasets to characterize the complexity of the respective classification problems. Their use in recent literature is also reviewed and discussed, allowing to prospect opportunities for future work in the area. Finally, descriptions are given on an R package named Extended Complexity Library (ECoL) that implements a set of complexity measures and is made publicly available.

How Complex Is Your Classification Problem?: A Survey on Measuring Classification Complexity

The classification in class imbalanced data has drawn significant interest in medical application. Most existing methods are prone to categorize the samples into the majority class, resulting in bias, in particular the insufficient identification of minority class. A kind of novel approach, class weights random forest is introduced to address the problem, by assigning individual weights for each class instead of a single weight. The validation test on UCI data sets demonstrates that for imbalanced medical data, the proposed method enhanced the overall performance of the classifier while producing high accuracy in identifying both majority and minority class.

Class Weights Random Forest Algorithm for Processing Class Imbalanced Medical Data

Documents containing text and graphics components are usually acquired as binary images for computer processing purposes. Salt-and-pepper noise is a prevalent artifact in such images. Removing this noise usually requires iterative or multiple-pass processing, some techniques even cause distortions in document components. In this paper, we propose a novel method based on the kFill algorithm that can be accomplished in single-pass scan over the image. The algorithm is capable of removing simultaneously both salt noise and pepper noise of any sizes that are smaller than the size of document objects. Results of the proposed method are given in comparison with the well-known morphological operations.

Removing salt-and-pepper noise in text/graphics images

The pathfinding analysis has importance for various working such as logistics, transportation, operation management, system analysis and design, project management, network and production line. Especially, game programming technology has effect to economic and dominates culture, increasingly. The shortest path analysis is artificial intelligent which developed capability about think cause and effect, learning and thinking like human. Heuristic technique is method for solving a problem which gets result or not. It may return unexpected value which depends on each problem. 3D game realtime strategy uses shortest path analysis control path for movement of character, wheeled vehicle, animals, and fantasy buildings. It uses path for movement ever time, both in sea, terra and sky (three layers). The position of 3D object use refers 2D coordinate. Generally, the most of 3D game strategy both on LAN and online will play on terrain more than inside building or the room. This research proposes two essences. It proposes comparison rapidity, intelligence and efficiency about seven path finding algorithms which is principle algorithm are familiar. These algorithms used in game computer extensively as follows [1] Depth First Search [2] Iterative Deepening [3] Breadth First Search [4] Dijkstra's Algorithm [5] Best First Search [6] A-Star Algorithm (A∗) [7] Iterative Deepening A∗. Besides, this research proposes Depth Direction A∗ algorithm method which new method which use linear graph theory cooperate with A∗ Algorithm calculate increase efficiency of avoid barrier object on maps and search for shortest path of multi-layer. It movement are natural characteristic more other method. The cost node grows less than A∗ algorithm in clear area and plan about expand node before processed. It guarantee about expand node less than A∗ algorithm node certainly. It takes time movement and avoid barrier object less than uses other algorithm. This work able decrease expand node 22.12–70.67% depend on property of area between pathway by define 3D rendering rate are stable.

Pathfinding of 2D & 3D game real-time strategy with depth direction A∗ algorithm for multi-layer

Improving classification rate constrained to imbalanced data between overlapped and non-overlapped regions by hybrid algorithms

This paper describes the development of an optical mark reader that can be used for counting the examination score from the multiple-choice answer sheet. The system is developed based on PC-type microcomputer connecting to an image scanner. The system operations can be distinguished into two modes: learning mode and operation mode. In the learning mode, the model corresponding to each type of answer sheet is constructed by extracting all significant horizontal and vertical lines in the blank-sheet image. Then, every possibly cross-line will be located to form rectangular area. In the operation mode, each sheet fed into the system has to be identified by matching the horizontal lines detected with every model. The data extraction from each area can be performed based on the horizontal and vertical projections of the histogram. For the answer checking purpose, the number of black pixels in each answer block is counted, and the difference of those numbers between the input and its corresponding model is used as decision criterion. Finally, the database containing a list of subjects, students, and scores can be created. The experimental results on many styles of answer sheets show the effectiveness of such a system.

Image-processing-oriented optical mark reader

Falls are significant public health problem In the last few years, several researches based on computer vision system have been developed to detect a person who has fallen to the ground This paper presents a novel fall detection technique namely the directional bounding box (DBB) to detect a falls event especially a situation of fall direction paralleling the line of camera's sight The DBB is constructed with perspective side view transformation of depth information Moreover, a new aspect ratio namely the center of gravity point (COG) is proposed to monitor human movement The proposed technique was evaluated with the video data set gathering from a RGB-D sensor The experimental result of the proposed technique was better both accuracy and response times than previous works

Krisana Chinnasarn

Papers

Removing salt-and-pepper noise in text/graphics images

Pathfinding of 2D & 3D game real-time strategy with depth direction A∗ algorithm for multi-layer

Improving classification rate constrained to imbalanced data between overlapped and non-overlapped regions by hybrid algorithms

Image-processing-oriented optical mark reader

Fall detection using directional bounding box