Synthetic Minority Oversampling Technique Method Research Articles

Malfunction diagnosis of main station of power metering system using LSTM-ResNet with SMOTE method

The power metering system is an important part of the smart grid for data acquisition and analysis. The fault state of the main station directly affects the stable and safe operation of the power metering system. Hinged on the real-world data supplied by the monitoring platform of the Metrology Center of Guangdong Power Grid Co., Ltd., we present a novel malfunction diagnosis method for the main station of the power metering system. The proposed method utilizes the synthetic mi-nority over-sampling technique (SMOTE) and designs a combined model of long short-term memory (LSTM) network and ResNet. SMOTE solves the sample imbalance problem. Furthermore, the combined LSTM-ResNet model employs LSTM to extract the time-dependent signal feature and exploits ResNet to optimize data flow. Consequently, the proposed LSTM-ResNet model improves training efficiency and malfunction diagnosis accuracy. The proposed diagnosis mthod is verifird on the real-world data, which proves the proposed method’s surpass traditional methods. A specific analysis of results and the practical application of the proposed method is also elaborated.

Comparison of Support Vector Machine Performance with Oversampling and Outlier Handling in Diabetic Disease Detection Classification

Diabetes mellitus is a disease that attacks chronic metabolism, characterized by the body’s inability to process carbohydrates, fats so that glucose levels are high. Diabetes mellitus is the sixth cause of death in the world. Classifying data about diabetes mellitus makes it easier to predict the disease. As technology develops, diabetes mellitus can be detected using machine learning methods. The method that can be done is the support vector machine. The advantage of SVM is that it is very effective in completing classification, so it can quickly separate each positive and negative point. This study aimed to obtain the best SVM classification model based on accuracy, sensitivity, and precision values in detecting diabetes by adding Synthetic Minority Over-Sampling Technique (SMOTE) and handling outliers. The SMOTE method was applied to handle class imbalance. The Support Vector Machine (SVM) method aimed to produce a function as a dividing line or what can be called a hyperplane that matches all input data with the smallest possible error. The data studied were indications of diabetes, consisting of 8-factor variables and 1 class variable. The test results show that the SVM-SMOTE scenario produces the best accuracy. The SVM SMOTE scenario produced an accuracy value of the RBF kernel of 88% with an error of 12%, and this is obtained from the division of test data and training data of 90:10. This SVM-SMOTE scenario produced a precision value of 0.880 and a sensitivity value of 0.880. The research results showed that factor classification was more accurate if it is carried out using the support vector machine (SVM) method with imbalance data handling (SMOTE), and it can be concluded that the distribution of test data and training data influences a test scenario.

Comparison of Machine Learning Methods for Classifying User Satisfaction Opinions of the PeduliLindungi Application

Since the emergence of the Covid-19 virus, the Indonesian government urged people to study, work, and worship or work from home. The social restriction policy has changed people's behavior which requires physical distance in social interaction. The government developed an application to minimize the spread of Covid-19, namely the PeduliLindungi application. The PeduliLindungi application is a tracking application to prevent the spread of Covid-19. The government's policy of implementing the PeduliLindungi application during Covid-19 aroused pros and cons from the public. The volume of PeduliLindungi application review data on Google Play was increasing, so manual analysis could not be done. New analytical approaches needed to be carried out, such as sentiment analysis. This research aimed to analyze user reviews of the PeduilLindungi application using classification methods, namely Support Vector Machine (SVM), Random Forest, and Naïve Bayes. The methods used were Synthetic Minority Oversampling Technique (SMOTE), Random Forest, SVM, and Naïve Bayes. SMOTE was used to balance user review data on the PeduliLindungi application. After the data had been balanced, classification was carried out. The results of this study showed that the Random Forest method with SMOTE got better accuracy than the SVM and Naive Bayes methods, which was 96.3% based on the division of training and testing data using 10-fold cross-validation. Thus, using the SMOTE method could improve the accuracy of classification methods in classifying opinions of user satisfaction with the PeduliLindungi application.

Enhanced classification of hydraulic testing of directional control valves with synthetic data generation

Production environments bring inherent system challenges that are reflected in the high-dimensional production data. The data is often nonstationary, is not available in sufficient size and quality, and is class imbalanced due to the predominance of good parts. Data-driven manufacturing analytics requires data of sufficient quantity and quality. In order to predict quality characteristics, production data is collected across processes in the industrial use case at Bosch Rexroth AG for the purpose of inferring results in hydraulic final inspection using machine learning methods. Since high quality data generation is costly, synthetic data generation methodologies offer a promising alternative to improve prediction models and thus generate safer, more accurate predictions for manufacturing companies. Among the synthetic data generation methodologies used, variational autoencoders compared to generative adversarial networks and synthetic minority oversampling technique methods are best suited to synthesize the feature with highest feature importance from a small sample data set compared to the production data and improve the prediction for the target variable.

New model combination meta-learner to improve accuracy prediction P2P lending with stacking ensemble learning

Peer-to-peer (P2P) Lending is a type of financial innovation that offers loans without intermediaries to individuals and companies. In the P2P lending system, there is a risk of default on the loan which causes the company to lose. Many studies have to reduce the risk of default by developing a classification model of prediction of default that focuses on increasing accuracy. However, the big problem with prediction is data imbalance and low performance classification algorithms. The purpose of this study is to improve the accuracy of default risk prediction by balancing the data and combining the stacking model ensemble with the meta-learner. The proposed new model consists of 3 optimization parts, the first is Synthetic Minority Oversampling Technique (SMOTE), the second is the selection of features and the third is stacking ensemble learning. The SMOTE method is used to balance the data, the feature selection LightGBM and stacking ensemble learning (LGBFS-StackingXGBoost) to optimize machine learning accuracy. A new model of stacking ensemble learning by combining three base-learner algorithms namely KNN, SVM and Random Forest into the XGBoost meta-learner algorithm. The model was tested using two datasets, namely the online P2P lending dataset and the lending club loan data analysis dataset. The evaluation results show that LGBFS-StackingXGBoost is the best model for both datasets. In the online P2P lending dataset, it received an accuracy of 99,982% and in the lending club loan data analysis dataset, it received an accuracy of 91,434%. This study shows that the accuracy of the prediction model can be improved using the LGBFS-StackingXGBoost method.

Estimation Chickpea Species and Productivity per Decare with Synthetic Data Generation Methods

Production increase in agriculture depends on some parameters such as improving arable land, activating spraying and irrigation activities. In addition to these, it is known that spraying and seed types have an effect on productivity. Therefore, proper selection of seed types is important. With the developing technology, big data consisting of scientific studies can be recorded digitally and used in the estimation or decision-making process. In this study, chickpea species diversity was made with classification process using machine learning methods by taking advantage of the characteristics of chickpea plant. In addition, productivity per decare was estimated by regression process. Accuracy was preferred as a success criterion for classification, and rmse success criterion was preferred for regression. The dataset was first used raw, and then experiments were made using synthetic data. To generate synthetic data, the synthetic minority oversampling technique method and also the n-shifting mean method proposed in this study were used. When the success rates of the results obtained were compared, the highest success rate was 90.6% in the classification made using only raw data. Likewise, the classification success rate of the dataset using the synthetic data created with the raw data was the highest 100%. For regression, the highest score was 0.17 for raw data and 0.16 for synthetic data. The high performance of the results showed that machine learning algorithms can be used in this field.

Machine learning's performance in classifying postmenopausal osteoporosis Thai patients

This work investigates the performance of different machine learning (ML) methods for classifying postmenopausal osteoporosis Thai patients. Our dataset contains 377 samples compiled retrospectively using the medical records of a Thai woman in the postmenopause stage from the obstetrics and gynecology clinic, Ramathibodi Hospital, Bangkok, Thailand. Missing data imputation, feature selection, and handling imbalanced techniques are independently applied as pre-processing approaches. The performance of different ML algorithms, including k-nearest neighbors (k-NN), neural network (NN), naïve Bayesian (NB), Bayesian network (BN), support vector machine (SVM), random forest (RF), and decision tree (DT), is compared between the pre-processed and original data. The results demonstrate that different ML algorithms combined with pre-processing techniques achieve varying results. In terms of accuracy, the three best-performing methods are the NN, NB, and RF models when a wrapper approach is used with an appropriate learner. In terms of specificity, the DT model achieves the best performance when the synthetic minority oversampling technique method is applied. When feature selection techniques are applied, the k-NN, BN, and SVM algorithms obtain the best sensitivity, whereas the NN shows the best area under the curve. Overall, in comparison with the original dataset, the pre-processed approaches improved model performance. Therefore, proper pre-processing techniques should be considered when developing ML classifiers to identify the best appropriate model.

Diagnosing Coronary Artery Disease on the Basis of Hard Ensemble Voting Optimization.

Background and Objectives: Recently, many studies have focused on the early diagnosis of coronary artery disease (CAD), which is one of the leading causes of cardiac-associated death worldwide. The effectiveness of the most important features influencing disease diagnosis determines the performance of machine learning systems that can allow for timely and accurate treatment. We performed a Hybrid ML framework based on hard ensemble voting optimization (HEVO) to classify patients with CAD using the Z-Alizadeh Sani dataset. All categorical features were converted to numerical forms, the synthetic minority oversampling technique (SMOTE) was employed to overcome imbalanced distribution between two classes in the dataset, and then, recursive feature elimination (RFE) with random forest (RF) was used to obtain the best subset of features. Materials and Methods: After solving the biased distribution in the CAD data set using the SMOTE method and finding the high correlation features that affected the classification of CAD patients. The performance of the proposed model was evaluated using grid search optimization, and the best hyperparameters were identified for developing four applications, namely, RF, AdaBoost, gradient-boosting, and extra trees based on an HEV classifier. Results: Five fold cross-validation experiments with the HEV classifier showed excellent prediction performance results with the 10 best balanced features obtained using SMOTE and feature selection. All evaluation metrics results reached > 98% with the HEV classifier, and the gradient-boosting model was the second best classification model with accuracy = 97% and F1-score = 98%. Conclusions: When compared to modern methods, the proposed method perform well in diagnosing coronary artery disease, and therefore, the proposed method can be used by medical personnel for supplementary therapy for timely, accurate, and efficient identification of CAD cases in suspected patients.

Developing predictive models for early detection of intervertebral disc degeneration risk

Intervertebral disc degeneration (IVDD) has been associated with causing chronic back pain. Lower back pain is a common cause of doctor’s visits and disability and affects 11.9% of the worldwide population. To the best of our knowledge, this work is the first to develop a tool for predicting the risk of IVDD using machine learning algorithms (MLAs) to aid in the earlier detection of the disease. MLAs utilized in this study include logistic regression, decision tree, artificial neural network, boosting, bagging, and random forest. The Synthetic Minority Oversampling Technique (SMOTE) method was performed on the unbalanced data set to up sample the minority class (patients with disc degeneration). The cluster-then-predict technique was also utilized to determine clusters in the data to be used as another risk factor or input variable into the classification algorithms. Results showed that the bagging algorithm developed the best-performing model. Results also demonstrated that the performance of the models improved after using the cluster-then-predict technique, which supports previous literature. Moreover, an improvement in the prediction of the minority class was also observed when the SMOTE method was applied. The finding from this study can be utilized to develop a predictive tool for the early detection of disc degeneration. This predictive tool can be used by doctors and physicians in the office with patients to closely monitor and assess disc degeneration risk.

Development and Validation of a Radiomics Nomogram for Predicting Clinically Significant Prostate Cancer in PI-RADS 3 Lesions.

PurposeTo develop and validate a radiomics nomogram for the prediction of clinically significant prostate cancer (CsPCa) in Prostate Imaging-Reporting and Data System (PI-RADS) category 3 lesions.MethodsWe retrospectively enrolled 306 patients within PI-RADS 3 lesion from January 2015 to July 2020 in institution 1; the enrolled patients were randomly divided into the training group (n = 199) and test group (n = 107). Radiomics features were extracted from T2-weighted imaging (T2WI), apparent diffusion coefficient (ADC) imaging, and dynamic contrast-enhanced (DCE) imaging. Synthetic minority oversampling technique (SMOTE) was used to address the class imbalance. The ANOVA and least absolute shrinkage and selection operator (LASSO) regression model were used for feature selection and radiomics signature building. Then, a radiomics score (Rad-score) was acquired. Combined with serum prostate-specific antigen density (PSAD) level, a multivariate logistic regression analysis was used to construct a radiomics nomogram. Receiver operating characteristic (ROC) curve analysis was used to evaluate radiomics signature and nomogram. The radiomics nomogram calibration and clinical usefulness were estimated through calibration curve and decision curve analysis (DCA). External validation was assessed, and the independent validation cohort contained 65 patients within PI-RADS 3 lesion from January 2020 to July 2021 in institution 2.ResultsA total of 75 (24.5%) and 16 (24.6%) patients had CsPCa in institution 1 and 2, respectively. The radiomics signature with SMOTE augmentation method had a higher area under the ROC curve (AUC) [0.840 (95% CI, 0.776–0.904)] than that without SMOTE method [0.730 (95% CI, 0.624–0.836), p = 0.08] in the test group and significantly increased in the external validation group [0.834 (95% CI, 0.709–0.959) vs. 0.718 (95% CI, 0.562–0.874), p = 0.017]. The radiomics nomogram showed good discrimination and calibration, with an AUC of 0.939 (95% CI, 0.913–0.965), 0.884 (95% CI, 0.831–0.937), and 0.907 (95% CI, 0.814–1) in the training, test, and external validation groups, respectively. The DCA demonstrated the clinical usefulness of radiomics nomogram.ConclusionThe radiomics nomogram that incorporates the MRI-based radiomics signature and PSAD can be conveniently used to individually predict CsPCa in patients within PI-RADS 3 lesion.

Efficiency Analysis and Estimation of Factors Affecting the Efficiency with Decision Trees in Imbalanced Data: A Case of Turkey’s Environmental Sustainability

Cities have proliferated and experienced increasing environmental issues in the modern world. The concept of environmental sustainability is one of the main problems to solve. Therefore, it is fundamental to establish statistical methods to measure environmental sustainability. The first aim of this study is to measure the environmental sustainability performance of 42 cities in Turkey by Data Envelopment Analysis. The second aim is to solve the imbalance in the efficiency values obtained using Synthetic Minority Oversampling Technique methods. After all, we expose the multiple relationships between input and output variables and efficiency using the Decision trees classifiers approach. As a result of the analyses, three internal factors were found to influence the environmental efficiency levels: residential sales, population intensity, and the number of completed industrial sites. It has been determined that the number of completed industrial sites and the increase in residential sales distorted environmental efficiency.

Study of thermal sensation prediction model based on support vector classification (SVC) algorithm with data preprocessing

In order to meet people's demand for comfort, indoor thermal environment often needs to be adjusted. Nevertheless, HVAC target parameters are often over-setted due to cognitive bias, resulting in an uncomfortable environment and waste of energy. Therefore, a thermal sensation prediction model is required to advise the setting of environmental and individual parameters. An accurate mode for predicting thermal sensation encountered severe challenges because of sensor errors, environmental noise, subjective thermal sensation differences, and sample data imbalance. The concrete aim of this paper is to verify the potential of using Support Vector Classification (SVC) algorithm to predict thermal sensation vote (TSV) based on Edited Nearest Neighbour (ENN) and Synthetic Minority Oversampling Technique (SMOTE), named combined ENN + SMOTE + SVC method. Firstly, for the problem of outliers in the dataset, the ENN method was adopted to clean the raw data. Secondly, SMOTE method is used to expand the training data which has the problem of sample imbalance. Finally, SVC algorithm is adopted to build a thermal sensation prediction model. The results show that the model built by the combined ENN + SMOTE + SVC method can achieve better performance than PMV index and other classic classification algorithms.

Memory-Efficient Deep Learning for Botnet Attack Detection in IoT Networks

Cyber attackers exploit a network of compromised computing devices, known as a botnet, to attack Internet-of-Things (IoT) networks. Recent research works have recommended the use of Deep Recurrent Neural Network (DRNN) for botnet attack detection in IoT networks. However, for high feature dimensionality in the training data, high network bandwidth and a large memory space will be needed to transmit and store the data, respectively in IoT back-end server or cloud platform for Deep Learning (DL). Furthermore, given highly imbalanced network traffic data, the DRNN model produces low classification performance in minority classes. In this paper, we exploit the joint advantages of Long Short-Term Memory Autoencoder (LAE), Synthetic Minority Oversampling Technique (SMOTE), and DRNN to develop a memory-efficient DL method, named LS-DRNN. The effectiveness of this method is evaluated with the Bot-IoT dataset. Results show that the LAE method reduced the dimensionality of network traffic features in the training set from 37 to 10, and this consequently reduced the memory space required for data storage by 86.49%. SMOTE method helped the LS-DRNN model to achieve high classification performance in minority classes, and the overall detection rate increased by 10.94%. Furthermore, the LS-DRNN model outperformed state-of-the-art models.

Classification of Metastatic and Non-Metastatic Thoracic Lymph Nodes in Lung Cancer Patients Based on Dielectric Properties Using Adaptive Probabilistic Neural Networks.

ObjectiveDielectric properties can be used in normal and malignant tissue identification, which requires an effective classifier because of the high throughput nature of the data. With easy training and fast convergence, probabilistic neural networks (PNNs) are widely applied in pattern classification problems. This study aims to propose a classifier to identify metastatic and non-metastatic thoracic lymph nodes in lung cancer patients based on dielectric properties.MethodsThe dielectric properties (permittivity and conductivity) of lymph nodes were measured using an open-ended coaxial probe. The Synthetic Minority Oversampling Technique method was adopted to modify the dataset. Feature parameters were scored to select the appropriate feature vector using a Statistical Dependency algorithm. The dataset was classified using adaptive PNNs with an optimized smooth factor using the simulated annealing PNN (SA-PNN). The results were compared with traditional Probabilistic, Support Vector Machines, k-Nearest Neighbor and the Classify functions in MATLAB.ResultsThe conductivity frequencies of 3959, 3958, 3960, 3978, 3510, 3889, 3888, and 3976 MHz were selected as the feature vectors for 219 lymph nodes (178 non-metastatic and 41 metastatic). Compared with the other methods, SA-PNN achieved the highest classification accuracy (92.92%) and the corresponding specificity and sensitivity were 94.72% and 91.11%, respectively.ConclusionsCompared with the other methods, the SA-PNN proposed in the present study achieved a higher classification accuracy, which provides a new scheme for classification of metastatic and non-metastatic thoracic lymph nodes in lung cancer patients based on dielectric properties.

A new approach for the detection of abnormal heart sound signals using TQWT, VMD and neural networks

Phonocardiogram (PCG) plays an important role in evaluating many cardiac abnormalities, such as the valvular heart disease, congestive heart failure and anatomical defects of the heart. However, effective cardiac auscultation requires trained physicians whose work is tough, laborious and subjective. The objective of this study is to develop an automatic classification method for anomaly (normal vs. abnormal) detection of PCG recordings without any segmentation of heart sound signals. Hybrid signal processing and artificial intelligence tools, including tunable Q-factor wavelet transform (TQWT), variational mode decomposition (VMD), phase space reconstruction (PSR) and neural networks, are utilized to extract representative features in order to model, identify and detect abnormal patterns in the dynamics of PCG system caused by heart disease. First, heart sound signal is decomposed into a set of frequency subbands with a number of decomposition levels by using the TQWT method. Second, VMD is employed to decompose the subband of the heart sound signal into different intrinsic modes, in which the first four intrinsic modes contain the majority of the heart sound signal’s energy and are considered to be the predominant intrinsic modes. They are selected to construct the reference variable for analysis. Third, phase space of the reference variable is reconstructed, in which the properties associated with the nonlinear PCG system dynamics are preserved. Three-dimensional PSR together with Euclidean distance has been utilized to derive features, which demonstrate significant difference in PCG system dynamics between normal and abnormal heart sound signals. Finally, PhysioNet/CinC Challenge heart sound database is used for evaluation and the synthetic minority over-sampling technique method is applied to balance the datasets. By using the 10-fold cross-validation style, experimental results demonstrate that the proposed features with dynamical neural networks based classifier yield classification performance with sensitivity, specificity, overall score and accuracy values of 97.73 $$\%$$ , 98.05 $$\%$$ , 97.89 $$\%$$ , and 97.89 $$\%$$ , respectively. The results verify the effectiveness of the proposed method which can serve as a potential candidate for the automatic anomaly detection in the clinical application.

CLASSIFICATION BOOSTING IN IMBALANCED DATA

Most existing classification approaches assumed underlying training data set to be evenly distributed. However, in the imbalanced classification, the training data set of one majority class could far surpass those of the minority class. This becomes a problem because it’s usually produces biased classifiers that have a higher predictive accuracy over the majority class, but poorer predictive accuracy over minority class. One popular method recently used to rectify this is the SMOTE (Synthetic Minority Over-Sampling Technique) which combines algorithms at data level. Therefore, this paper presents a novel approach for learning and imbalanced data sets, based on a combination of the SMOTE algorithm and the boosting procedure by focusing on a two-class problem. The Bidikmisi data set is imbalanced, because the distribution of majority class examples is 15 times the number of minority class examples. All models have been evaluated using stratified 5-fold cross-validation, and the performance criteria (such as Recall, F-Value and G-Mean) are examined. The results show that the SMOTE-Boosting algorithms have a better classification performance than the AdaBoost.M2 method, as the g-mean value increases 4-fold after the SMOTE method is used. We can say that SMOTE-Boosting algorithm is quite successful when taking advantage of boosting algorithms with SMOTE. When boosting affects the accuracy of the random forest by focusing on all data classes, the SMOTE algorithm alters the performance values of the random forest only in minority classes.

ACHILLES: Accuracy-Aware High-Level Synthesis Considering Online Quality Management

In this paper, we present an accuracy-aware design framework [called accuracy-aware high-level synthesis (Achilles)], which synthesizes a high-level description of an input application with the objective of minimizing the energy consumption of the synthesized circuit. The proposed framework includes two main parts of Achilles and light-weight predictor selection. The framework leverages light-weight error predictors (i.e., machine learning-based classifiers) to achieve more energy reduction by dynamically managing the output quality level (exact or approximate) of the synthesized circuit. To synthesize the input application, first, we exploit a heuristic algorithm to determine the quality level required for each operation in the data flow graph (DFG) representation of the input application. Next, for synthesizing the input application, we propose an effective Achilles algorithm which utilizes the flexibility of the available multiquality arithmetic units in a high-level cell library to synthesize the datapath. To improve the efficiency, the process starts by iteratively reducing the number of functional units required for synthesizing the DFG. Then, a proper light-weight error predictor satisfying the user expected quality is chosen from the available predictors in the framework. Based on the quality requirements, three different quality management modes are considered. The efficacy of the proposed framework is assessed for benchmarks from image and signal processing as well as robotics domains. The study of these benchmarks indicates that Achilles may reduce the energy consumption up to 51% (36% on average), up to 72% (51% on average), and up to 57% (33% on average) in threshold, average, and hybrid modes, respectively, for the studied cases. Moreover, the results show that relative coverage of large errors may be increased from 21% to 55% by employing synthetic minority oversampling technique method.

Dynamic Synthetic Minority Over-Sampling Technique-Based Rotation Forest for the Classification of Imbalanced Hyperspectral Data

Rotation forest (RoF) is a powerful ensemble classifier and has attracted substantial attention due to its performance in hyperspectral data classification. Multi-class imbalance learning is one of the biggest challenges in machine learning and remote sensing. The standard technique for constructing RoF ensemble tends to increase the overall accuracy; RoF has difficulty to sufficiently recognize the minority class. This paper proposes a novel dynamic SMOTE (synthetic minority oversampling technique)-based RoF algorithm for the multi-class imbalance problem. The main idea of the proposed method is to dynamically balance the class distribution before building each rotation decision tree. A resampling rate is set in each iteration (ranging from 10% in the first iteration to 100% in the last) and this ratio defines the number of minority class instances randomly resampled (with replacement) from the original dataset in each iteration. The rest of the minority class instances are generated by the SMOTE method. The reported results on three real hyperspectral datasets show that the proposed method can get better performance than random forest, RoF, and some popular data sampling methods.

Prediction of Default Probability by using Statistical Models for rare Events

Summary Prediction models in credit scoring usually involve the use of data sets with highly imbalanced distributions of the event of interest (default). Logistic regression, which is widely used to estimate the probability of default, PD, often suffers from the problem of separation when the event of interest is rare and consequently poor predictive performance of the minority class in small samples. A common solution is to discard majority class examples, to duplicate minority class examples or to use a combination of both to balance the data. These methods may overfit data. It is unclear how penalized regression models such as Firth’s estimator, which reduces bias and mean-square error relative to classical logistic regression, performs in modelling PD. We review some methods for class imbalanced data and compare them in a simulation study using the Taiwan credit card data. We emphasize the effect of events per variable for developing an accurate model—an often neglected concept in PD-modelling. The data balancing techniques that are considered are the random oversampling examples and synthetic minority oversampling technique methods. The results indicate that the synthetic minority oversampling technique improved predictive accuracy of PD regardless of sample size. Among the penalized regression models that are analysed, the log-F prior and ridge regression methods are preferred.

Computed Tomography-Based Radiomic Features Could Potentially Predict Microsatellite Instability Status in Stage II Colorectal Cancer: A Preliminary Study

To investigate whether quantitative radiomics features extracted from computed tomography (CT) can predict microsatellite instability (MSI) status in an Asian cohort of patients with stage Ⅱ colorectal cancer (CRC). This retrospective study was approved by our institutional review board, and the informed consent requirement was waived. From March 2016 to March 2018, 119 Chinese patients with pathologically confirmed stage Ⅱ CRC, available MSI status, and preoperative contrast-enhanced CT images were included in this study. Clinical and pathological information was obtained from the institutional database. The radiomics features were extracted from portal venous-phase CT images of segmented volumes of each entire primary tumor by using Matrix Laboratory (MATLAB), and radiomics signatures were generated using the least absolute shrinkage and selection operator logistic regression model. The minority group was balanced via synthetic minority over-sampling technique method. The association between the clinicopathologic characteristics and MSI status was assessed using Student's t test, Chi-square, or Fisher's exact test. The predictive efficacy of MSI status using radiomics features, clinical factors (including age, gender, CT-reported tumor location, differentiation degree of tumor, smoking history, hypertension history, family history of cancer, diabetes history, level of the Ki-67 expression, and laboratory analysis) and the combined models were evaluated, respectively. Predictive performance was evaluated by the area under receiver operating characteristic curve, accuracy, sensitivity, and specificity. MSI status was significantly associated with tumor location (p = 0.043); differentiation degree of tumor (p < 0.0001), hypertension history (p = 0.012), and the level of the Ki-67 expression (p = 0.015). Six radiomics features and 11 clinical characteristics were selected for predicting MSI status. The model that used the combination of clinical factors and radiomics features achieved the overall best performance than using either of the two features alone, yielding the area under the curve, sensitivity, and specificity of 0.752, 0.663, 0.841 for the combined model, 0.598, 0.371, 0.825 for clinical factors alone, and 0.688, 0.517, 0.858 for radiomics features alone, respectively. CT-based radiomic features of stage Ⅱ CRC are associated with MSI status. Combining analysis of clinical features and CT features could improve predictive efficacy and could potentially select the patients for individualized therapy noninvasively.