• <tr id="yyy80"></tr>
  • <sup id="yyy80"></sup>
  • <tfoot id="yyy80"><noscript id="yyy80"></noscript></tfoot>
  • 99热精品在线国产_美女午夜性视频免费_国产精品国产高清国产av_av欧美777_自拍偷自拍亚洲精品老妇_亚洲熟女精品中文字幕_www日本黄色视频网_国产精品野战在线观看 ?

    LDSVM:Leukemia Cancer Classification Using Machine Learning

    2022-08-24 03:31:16AbdulKarimAzhariAzhariMobeenShahrozSamirBrahimBelhaouriandKhabibMustofa
    Computers Materials&Continua 2022年5期

    Abdul Karim,Azhari Azhari,*,Mobeen Shahroz,Samir Brahim Belhaouri and Khabib Mustofa

    1Department of Computer Science&Electronics,Universitas Gadjah Mada,Yogyakarta,55281,Indonesia

    2Department of Artificial Intelligence,The Islamia University of Bahawalpur,Bahawalpur,63100,Pakistan

    3Division of Information&Computer Technology,College of Science&Engineering,Hamad Bin Khalifa University,Doha,5825,Qatar

    Abstract: Leukemia is blood cancer, including bone marrow and lymphatic tissues,typically involving white blood cells.Leukemia produces an abnormal amount of white blood cells compared to normal blood.Deoxyribonucleic acid (DNA) microarrays provide reliable medical diagnostic services to help more patients find the proposed treatment for infections.DNA microarrays are also known as biochips that consist of microscopic DNA spots attached to a solid glass surface.Currently, it is difficult to classify cancers using microarray data.Nearly many data mining techniques have failed because of the small sample size, which has become more critical for organizations.However, they are not highly effective in improving results and are frequently employed by doctors for cancer diagnosis.This study proposes a novel method using machine learning algorithms based on microarrays of leukemia GSE9476 cells.The main aim was to predict the initial leukemia disease.Machine learning algorithms such as decision tree (DT), naive bayes (NB),random forest (RF), gradient boosting machine (GBM), linear regression(LinR), support vector machine (SVM), and novel approach based on the combination of Logistic Regression(LR),DT and SVM named as ensemble LDSVM model.The k-fold cross-validation and grid search optimization methods were used with the LDSVM model to classify leukemia in patients and comparatively analyze their impacts.The proposed approach evaluated better accuracy, precision, recall, and f1 scores than the other algorithms.Furthermore, the results were relatively assessed, which showed LDSVM performance.This study aims to successfully predict leukemia in patients and enhance prediction accuracy in minimum time.Moreover, a Synthetic minority oversampling technique(SMOTE)and Principal compenent analysis(PCA)approaches were implemented.This makes the records generalized and evaluates the outcomes well.PCA reduces the feature count without losing any information and deals with class imbalanced datasets,as well as faster model execution along with less computation cost.In this study, a novel process was used to reduce the column results to develop a faster and more rapid experiment execution.

    Keywords: Leukemia; GSE9476; cancer; genes; classification; machine learning;ensemble LDSVM classifier

    1 Introduction

    Leukemia is the most common type of blood cancer in all age groups, particularly in children.This abnormal concept is caused by increased blood cell proliferation and immature growth,which can harm red blood cells,brain tissue,and the immune system.A cell is instructed by the genetic code when to reproduce and when to die.For cancer to occur,changes in gene expression may lead to defective instructions.The production of proteins is also affected by cells because they are under the control of cells[1].Some genes modify proteins to fix damaged cells,which may be related to cancer.If parents have these mutations, they may be inherited by their offspring.There are some genetic variations at birth, and the risk increases after lifestyle factors, such as smoking and exposure to the sun.Chemical signaling changes contribute to cancer,which dictates how the body uses the genetic material deoxyribonucleic acid(DNA).For certain cancers,a person may inherit a physical predisposition.The most significant step towards drug design and discovery is the detection of a disease that leads to the development of pharmaceutical therapies.The primary goal of gene identification and classification of the genome’s diseased genes is to identify any non-disease genes.The diagnosis of a disease is critical in medicine and pharmacology to determine the best treatment method.Observing the genes associated with the disease is the first step in coping with inherent disorders.Genetic abnormalities arise because of anomalies in single genes or gene regions.Because of the numerous genes,classifying disease-related genes is a complicated task in machine learning.

    Cancer is a widely defined concept that a degenerative disease occurs when cellular changes induce the uncontrolled growth and division of cells.Some cancers cause cells to overgrow,while other cancers cause cells to develop and split more rapidly.Some cancer types may result in noticeable growths called tumors, while others do not, such as leukemia.Most hematologic cancers start in the bone marrow.Blood cancers occur when abnormal blood cells start growing out of control, interrupting normal blood cell function,fighting off infection,and producing new blood cells.There are about a hundred different types of cancer.The three main types of blood cancers are leukemia,myeloma,and lymphoma.Blood and bone marrow cancers are called leukemias,which begin in the blood-forming tissue of the bone marrow.These forms of cancer are not solid tumors that develop[2].Instead,large numbers of dysfunctional white blood cells(leukemia cells and leukemia blast cells)become so crowded that they do not have the space to live in the blood and bone marrow.Low red blood cells can make it more difficult for the body to obtain the oxygen required into its tissues, regulate blood loss, or combat infections.The four types of leukemia vary from each other,grouped according to how rapidly cancers evolve(acute or chronic)and how leukemia develops,which type of blood cells it begins with(lymphoblastic or myeloid).Acute myelogenous leukemia(AML),acute lymphocytic leukemia(ALL),chronic myelogenous leukemia(CML),and chronic lymphocytic leukemia(CLL)are the four major forms of leukemia.

    Machine learning algorithms are techniques used for the identification and treatment of leukemia.This study represents the decision tree(DT),support vector machine(SVM),linear regression(LinR),naive Bayes(NB),random forest(RF),gradient boosting classifiers(GBM),and proposed ensemble LDSVM(LR+DT+SVM)machine learning algorithms used to identify leukemia in its subtypes,and presents a comparative analysis of these algorithms[3].The DNA microarrays of the Leukemia GSE9476 cell-based dataset were collected from the Kaggle benchmark dataset repository.This dataset consists of 22285 genes associated with five different leukemia types, namely bone_marrow_CD34,bone marrow, AML, PB, and PBSC_CD34.The feature vectors of the dataset were split into two partitions, training data, and testing data, with ratios of 80% and 20%.Various machine learning algorithms have been applied to the training and prediction processes,and different experiments are performed by tuning the hyperparameters.The structure of the experimental prototype is designed based on the assumption that the proposed approach can be easily implemented with any other dataset without any manual setups.The focus is on the development of a portable,exible,and efficient method for cancer classification.The ensemble LDSVM model presents promising results that outperform other machine-learning algorithms[4].This model is embedded with k-fold cross-validation,and grid search optimization techniques, which makes it more efficient and improves the accuracy results.The performance of the proposed novel classification approach for leukemia cancer is based on the composition of multiple components, such as k-fold cross-validation, grid-search optimization,and ensemble LDSVM which are evaluated through different scientific measures such as accuracy,precision,recall,and F1-score.The classification of leukemia cancer based on the proposed approach,is shown in Fig.1.The contributions of this study areas follows.

    Figure 1:Classification of leukemia cancer based on proposed approach

    ·The proposed study aims to classify leukemia cancer in patients with the highest prediction accuracy and quick response to help doctors treat patients properly and save their lives.

    ·The microarray of DNA genes was used as a dataset that consisted of 22285 genes associated with leukemia cancer patients and seven different machine learning models were applied to carry out the experiments with the Leukemia dataset to analyze the effects.

    ·The proposed approach consists of an ensemble LDSVM model that has its own novelty because of its structure with k-fold cross-validation and grid search optimization,as shown in Fig.5.A novel LDSVM ensemble model is proposed in this study to enhance the performance and prediction accuracy.

    ·The proposed approach and the effects of machine learning algorithms were analyzed using evaluation parameters such as accuracy,precision,recall,and F1-score.Furthermore,comparative analyses were performed to illustrate the transparent and promising results of the proposed approach.

    Principal component analysis (PCA) is a technique for performing adaptive data analysis.Moreover,it is adaptive in another sense,in that several variants of the technique have been developed and adopted to diverse data categories and structures.The synthetic minority oversampling technique(SMOTE) strategy used to deal with imbalanced datasets is to oversample the minority class in the first place.This is a type of data augmentation method.The simplest technique consists of duplicating examples from the minority class, despite the fact that these examples do not provide any new information to the model.Instead, new examples can be created by combining the elements of the current example.SMOTE and PCA techniques were employed for the generalization of records and predictions and for coping with class imbalanced datasets.The remainder of this paper is organized as follows.The relevant works addressed in this domain are presented in Section 2.The methodologies used in Section 3’s experimental outcomes and interpretation are analyzed in Sections 4 and 5, and the paper is concluded in Section 6.

    2 Related Work

    Data mining plays an essential role in predicting various types of Infections.Microarray technology provides recent advances in the capacity to calculate thousands of gene expressions at the exact moment.Such data analysis allows us to detect various clinical findings triggered by a few predictive genes countenance.The extraction and description of functions are conducted with the high precision of ensemble-based algorithms and a single understanding of linear regression.These allow exact rules to be extracted by describing gene expression differences between significantly expressed leukemia genes[5].Different classification algorithms for comparison criteria such as being regulated or unmonitored, if they are operating on tiny or massive databases, binary classifier or not if huge numbers will function the dimensions can be linearly divided and obtain precision.Leukemia is a disease that affects the bone marrow and blood.Manual processing of information is time-consuming and not precise.There have been attempts to build partial/fully automated cell segmentation and classification-based systems,but they are still in the demonstration process.Several existing mechanical systems collect sub-image functionality[6].In a study of microarray expression data from patients with systemic lupus erythematosus and PAPS,two autoimmune disorders of unknown genetic origin have several standard features.They employed an innovative statistical method for gene selection based on machine learning approaches,which included a combination of three policy discretization steps,a system of consensus selection,and multi-variable measurements of correlations.Different methods are available for the classification of cancer based on gene selection using microarray datasets.However,most of these studies did not consider detecting the minimum number of genes with high classification accuracy[7].

    A voting classifier model was suggested by [8] to improve classification accuracy by integrating several classifiers’predictions.Using CART, CHAID, and QUEST classifications, the performance measure gains precision,accuracy,specificity,and sensitivity measured for handling ionosphere data.The experimental results concluded that 93.84 percent of the test data selection achieved complete accuracy for the ensemble model with the choice of functions.The authors [9] collected oral data from different diagnostic centers that housed patients with and without cancer.The data obtained and information were preprocessed for replication,and the details are incomplete.They then used various algorithms for the classification of the NMDS dataset and analyzed the algorithms.The obtained classification rate was 100% for the C4.5 algorithm and 98.7% for the random forest algorithm.A study on an automated cell count was used in the methodology of the research[9].The explanatory method for automatic segmentation of white blood cells is discussed here.Next, the blood smear a priori data images often come from a simple threshold approaching the algorithm’s law.To obtain critical results,the marks were modified.This technique utilizes details of the composition of blood cells.In contrast to the ancient methods of using local context data, this technique is also valuable.When they perform proper segmentation,they have sharp boundaries for white blood cells.

    Chandrasekar et al.submitted actual classification technical.After reviewing various classifications, algorithms have selected six simulation-based classifiers in which the tree random classifier reaches an accuracy gain of 98 percent, which is a highly significant classification accuracy [10].Compared with various supervised algorithms, the best algorithms for classifiers were predicted.Experimental studies proved the viability of the proposed method.The template was tested with precision and recall.It was discovered that all other high-accuracy algorithms outperformed the random trees of different classification techniques.The nature of critical life-threatening clinical investigations, like breast [11] cancer, has been defined by an effective classifier.Voting classifiers can easily understand the approach and strategy for classification.Each element of the classification“votes”in an unweighted voting classifier community membership according to an uncertain sample and the feature value of which group is nearest.Voting by plurality/more wins.Voting classifiers were used and tested sporadically for application in gene expression data.

    Ancona et al.[12]discuss colon cancer gene tumors in their research.These findings indicate that the voting categories generate the best results for another individual classifier.Classification of cancer implies model building on datasets, including datasets for the expression of microarray genes and then differentiating the class value into each case in the sample by generating the model.Diagnostic findings may enable doctors to follow the appropriate therapy protocol for patients, particularly in the early diagnosis and treatment period.Many classification techniques have recently been developed for machine learning and have been introduced in cancer classification[13].Survival increases from 56 percent to more with an early and correct cancer diagnosis,decreasing the death rates by 86 percent in early diagnosis and treatment[14].Therefore,an effective and reliable system is needed for the early diagnosis of cancer.The microarray data set’s analysis and classification method have firmly proven that they provide an efficient way to diagnose diseases effectively,particularly in cancers.

    Furthermore,the microarray may be used to calculate at once in a cell assortment expression levels of a significant number of genes.Finally, the performance of the microarray was determined using microarray data.Data from gene expression is another popular micro-array data name.The word“ensemble learning”is used to generate a cooperative procedure using algorithms and tools in the computer and other fields.Multiple learners were more successful than individual learners.Together,learning can increase flexibility and improve outcomes in many areas, such as disease diagnostics,finance,bioinformatics,healthcare,factory development,and geography[15].By integrating classifiers using ensemble learning techniques, the precision and diversity of each classification can improve the generalization capability.Ensemble learning is also an excellent way to address the issue of class imbalances [16].In addition, a previous study [17] suggested.All these classifiers perform well in microarray data and come from various algorithm classifications, which means that they belong to multiple practices, that is, instability, probabilism, and stability.For example, the SVM algorithm is used to identify blood sampling as acute lymphoblastic Leukemia and acute myeloid Leukemia in the lymphoid stem and myeloid stem cells.The accuracy achieved by SVM[18]was 92%.The results show that logistic regression(LR)is a perfect algorithm compared to other algorithms used in classification and projection datasets.This experiment,and therefore must be noted that to our knowledge,this is the first time this dataset is used to classify logistic regression after the data set has been preprocessed,with an optimum speed of precision,accuracy,exactness,retrieval,and F1 ranking[19].

    3 Materials and Methods

    The leukemia cancer classification was proposed using the LDSVM classifier.The aim of this study was successfully achieved,which consists of the designed methodology and relative algorithms.This section describes the dataset used for the classification algorithms and proposed methods.

    3.1 Leukemia Cancer Dataset

    The dataset was obtained from a famous benchmark dataset repository Kaggle based on the Gene Expression Omnibus (GEO) dataset associated with the Curated Microarray Database (CuMiDa)repository containing 78 hand-picked cancer micro-arrange datasets and widely curated from over 30,000 studies.CuMiDa aims to provide homogeneous and advanced biological preprocessed forms of these datasets along with numerous 3-fold benchmark validation findings to propel machine studies aimed at the investigation of cancer.

    Fig.2 demonstrates the genes dataset presentation on five different variables, AML variable is higher in comparison with other features.Other features do not provide adequate partition, or the functionality of certain mechanisms is limited to particular leukemia cancers.The leukemia cancer based dataset was used with the code name GSE9476.The DNA gene microarray dataset consists of 22285 genes and five different leukemia types such as Bone_Marrow_CD34, Bone_Marrow, AML,PB, and PBSC_CD34.Fig.3 presents samples of genes associated with a dataset that illustrates the different gene variations in DNA microarrays.

    Figure 2:Leukemia GSE9476 genes dataset presentation based on five different types

    Figure 3:Variations of different gene’s data presentation based on classes

    3.2 Classification Algorithms Used in Leukemia

    Classification separates the data samples into target groups.For each data point,the classification technique provides the target class.Data classification is a controlled learning approach that establishes the class category[20].The dataset was divided into training and testing datasets.Classifiers were used to train the classification accuracy based on the testing dataset.Classification in the medical field is one of the most effective data-mining methods.However, the prediction accuracy of these methods varies depending on the classification algorithm used.It is challenging to find the best algorithm for classification from all available.This study proposes an exhaustive study of various classification algorithms.The assessment success with leukemia dataset for accuracy employed different classification methods to analyze microarray data such as DT,NB,RF,GBM,LinR,SVM,and ensemble LDSVM(LR+DT+SVM)approach[21].

    In Support Vector Machine (SVM), a hyperplane divides the entire space of the dataset by maximizing the class margin.A classifier training algorithm seeks the help vectors [22].As a nonparametric technique,SVM regression relies on a set of mathematical functions.The kernel converts the 2D data entries into a 3D shape to efficiently classify and handle the outliers.SVM solves regression problems with a linear function and maps the input vectors to an n-dimensional space called a function space when dealing with nonlinear regression problems.

    One of the most common and powerful data-mining techniques is the decision tree(DT).Many researchers have developed and tested this methodology well.In addition,data errors are also found in the learning process.Thus,it is crucial to work with large data volumes to achieve a decision tree algorithm that can generate a simple tree structure with high classification accuracy.In this study,we selected a dataset from Kaggle to run the decision tree algorithms.The choice of strategic division has a significant effect on the accuracy of a tree.The decision criteria are different for the classification and regression.

    The entropy is expressed mathematically for 1 attribute by:

    Entropy is expressed mathematically for several attributes as:

    IG is defined mathematically by:

    Random Forest (RF) is a tree-based ensemble model that makes predictions highly precisely based on the combination of several decision trees [23].The bagging technique is employed in this model to train several decision trees on different samples of bootstraps [24].In the RF sample, the training dataset with substitution is sub-sampled and is similar in size to the sample in the training dataset[25].In their prediction processes,random forest and other classifiers use the same techniques for building decision-making bodies.Defining the attributes of the root node at each stage is a significant challenge in their development.

    Logistic regression(LR)is an approach to mathematical knowledge processing in which a solution is found using one or more variables.Logistic regression is the type of regression used to predict class members’probabilities because if the target variable is categorical,it is the best learning model to use.LR processes the relationship between the group-dependent variable and one or more independent variables by calculating probabilities using a logistic function.

    The voting classifier model [26] comprises individually trained classifiers whose forecasts are combined when new data is labeled.The voting classifier incorporates a single composite classification to produce a few classifiers generated by weak learners.Thus, the error of every more inefficient learning algorithm can be reduced.Combining these classifiers aims to construct a voting classifier model that enhances the classification accuracy for individual classification.Combining the three models involves a weighted voting scheme [27] in which weighting is based on each prediction’s confidence value.The weights were then added, and the value was chosen with the highest number.The final selection of trust is the sum of the weights for the winning values divided by the number of models in the voting classifier.If a model does not predict greater confidence than the two combined predictions, no one will win.This ensemble of models increases classification accuracy and reduces training opportunities, as it prevents a partial decision by integrating various forecasts from the individual classifier.This study’s ensemble model combines the predictions of LR, SVM, and DT classifiers,and is then named LDSVM.

    4 Methodology

    Machine learning is a computer algorithm that automatically develops by means of practice.To make predictions without being programmed directly, machine learning algorithms construct a mathematical model based on sampling data known as“workout data.”Machine learning is closely related to the use of machine statistics to make predictions.Doctors have tools and machines to detect cancer to treat patients well.They have X-rays,MRI,and CT scan for cancer detection in the medical field,but these are all based on image scanning techniques and ready solid objects in the body.Blood analyses were performed by testing the blood samples through a laboratory tool,microarray,and DNA microarray.The microscope slide was used to hold the blood DNA genes in thousands of spots for testing.The doctor needs years of experience to analyze these tests.It takes years of training to make the correct prediction to treat the patient and save life.In other words, there is always a significant problem in predicting the right type of cancer in patients.The proposed study aims to solve this problem by proposing machine learning models and comprehensively studying their effects.

    In this study,different machine learning algorithms and voting classifiers have a better result of accuracy.Fig.4 shows the steps performed for better accuracy.After fetching, data pre-processing was performed.In the next step,different preprocessing techniques are used.For example,shuffling was conducted,and then the null values were removed smoothly to deal with the data categorically.In addition,punctuation was removed from the text during pre-processing.After preprocessing,the dataset was split into two subsets for training and testing.The ratios of the dataset division were 80%and 20%.After breaking the data,machine learning models were trained using essential features,and voting classifier models were used.The voting classifier models were LR, SVM, and DT.Machine learning models are tuned using different parameters.After model training, test data were passed to the trained models to evaluate the learning model’s performance and after this, other evaluation parameters were used, such as accuracy, F1-score, recall, and precision.This research collected a dataset of the maximum number of genes so that our model can be trained well on it and increase our accuracy by training our model.This model could detect cancer and treat patients in their cases well and treat them well.

    Figure 4:Leukemia cancer classification based on machine learning models and ensemble LDSVM(LR+DT+SVM)

    To prevent any problem,we perform shuffling on our data class-wise,so our dataset was the mix,and when we divide data into training and testing, we get good accuracy.The primary purpose of performing shuffling is to improve the performance of our model.Features already exist in terms of our genes.The dataset is already in numeric form, so no need to change it.The class data are in categorical form, but they require numeric form data to train our model.Here, the conversion was performed on categorical data and given a number for every class (e.g., bone marrow CD34 = 0).After this,it converts different classes into numeric forms,and our model works on it.All data were converted in numeric form.Using a novel approach, seven different machine learning models were applied to achieve the highest accuracy results.The results showed that the hybrid model had the best accuracy.

    4.1 Proposed LDSVM Architecture

    The LDSVM classification technique consists of k-fold cross-validation,grid search optimization,and an ensemble of models through voting.Fig.5 presents the complete architecture of the proposed approach.Three machine learning models were adopted to achieve the best accuracy results with the embedding of k-fold cross-validation and grid search optimization.The major challenges in machine learning are flexibility, portability, and efficiency.The 5 folds cross validation splits the data into a defined fold that is 5 for the experiments in which each split contains five folds, which means five divisions.K-fold cross-validation gives the best scores by selecting one by every single division as testing data and others as training to choose the efficient division.The experiments were carried out with n_spliters=5,n_repeats=3,and random_state=1 hyperparameters for 5-fold cross validation.It automates the data-split ion process for training.After this,the training data is passed to the machine learning model through grid search optimization which automates the hyperparameter tuning process of the models.A grid search was applied to each model and efficient hyperparameter values were selected for the efficient training of the models.

    Figure 5:Structure of proposed approach based on ensemble LDSVM model

    The logistic regression is implemented by using c={100,10,1.0,0.1,0.01},solver={‘newton-cg’,‘lbfgs’,‘liblinear’,‘sag’,‘saga’},max_iter={-1,100,150,200,400}and penalty={‘none’,‘l1’,‘l2’,‘elasticnet’}.The highest accuracy achieved with these{‘C’:10,‘penalty’:‘l2’,‘max_iter’:200,‘solver’:‘lbfgs’} values for logistic regression model.The SVM experimented by using c = {100, 10, 1.0, 0.1,0.01},kernel={‘linear’,‘poly’,‘rbf’,‘sigmoid’,‘precomputed’},max_iter={-1,100,150,200,400}and gemma={‘a(chǎn)uto’,‘scale’}.The best results were obtained using{c=1.0,kernel=‘linear’,max_iter= 200, and gemma = ‘scale’} hyperparameter values through a grid search.The DT model is used criterion={‘gini’,‘entropy’},splitter={“best,”“random”}max_depth={10,20,30,50,70,100}and min_sample_split={2,4,5,8,10}.The highest prediction accuracy was obtained using{criterion=‘entropy’,splitter=‘random’,max_depth=30,and min_sample_split=2}for the DT model.These algorithms are tuned using a grid search optimization technique.After the tuning and achieved the best accuracy results by grid search optimization with each model,voting was adopted to achieve even higher accuracy results,making it more efficient for the identification of leukemia cancer patients.

    4.2 Evaluation Criteria

    After all previous steps,Evaluation criteria were performed.This qualitative research used several evaluation metrics:accuracy, f1-score, recall, and precision.These evaluation criteria were used to evaluate machine learning models and ensemble classifiers.

    4.2.1 Accuracy

    Accuracy is the share of the total number of accurate predictions.It shall be decided by:

    where, TP = positives correctly classified/total positives, FP = negatives incorrectly classified/total negatives FP=false positive values and FN=false negative rate.

    4.2.2 Precision

    Precision is the proportion of the predicted positives as calculated using Eq.8.

    4.2.3 Recall

    It is the percentage of positive cases that were correctly identified,as calculated using the formula.

    4.2.4 F1-Score

    The F1-Score is an average performance metric that computes information retrieval precision using recall metrics.

    5 Results and Discussion

    The results of the experiments on leukemia cancer classification are discussed in this section.Numerous machine learning methods have been used to improve the prediction accuracy of leukemia cancer classification.Novel machine learning classification approaches were adopted to develop a leukemia detection system for GSE9476 cells based on microarray data.Several machine learning methods are used, such as DT, NB, RF, GBM, LR, SVM, and the proposed ensemble LDSVM,to achieve the highest prediction results.Three types of machine learning algorithms, LR, DT, and SVM,were used as ensemble classifiers for the proposed approach,known as LDSVM.An analysis of the proposed LDSVM and other ML models utilizing the confusion matrix is shown in Fig.6 by comparing the predicted and actual labels.The test data were used for evaluation.The prediction was combined as an ensemble classifier to obtain accurate results.For classification,ensemble classifiers were applied with soft and hard voting.Thus,the output shows 99.8%accuracy,using a soft voting hybrid approach with three models(LR,DT,and SVM)named LDSVM;with hard voting it reached 95.0%accuracy and 99.9%with 5 fold cross validation,grid search optimization,and LDSVM model.

    The evaluation results of ensemble classifiers with evaluation parameters to improve the system’s learning capabilities.Tab.1 shows the highest accuracy achieved during the evaluations and different models with different parameters.The proposed approach with the ensemble LDSVM model achieved an accuracy of up to 99%.The results show that the accuracy improves when using the ensemble learning approach.The ensemble method used in this study is based on a confusion matrix.A confusion matrix was used to calculate classifier accuracy.Different performance parameters such as accuracy,precision,recall,and f1score result are used with different models,and LDSVM achieves a better performance.The highest accuracy was obtained for LDSVM.The LDSVM performed better than the individual classifiers.

    Figure 6:Comparative analyses of machine learning model by using confusion matrix

    Table 1:Comparative results of machine learning models and proposed ensemble LDSVM

    Fig.7 shows the result of different performance parameters using other machine learning models.The best machine learning algorithms such as decision tree, random forest, na?ve Bayes, gradient boosting, linear regression, and SVM.The training set is sent to all selected machine learning algorithms, and the validation set is sent to the output of the classifier.Owing to the effectiveness of the LDSVM, an accuracy of 99% was achieved.However, in the context of leukemia cancer, it is more compatible to work with overfit data, and run time is very fast to process leukemia cancer,and many issues can be solved with the LDSVM.The results show that the LDSVM in the proposed approach with soft voting achieves 99.9%accuracy.Tab.1 shows a comparison of the prediction time as well.The least time is 0.223 s obtained with the GBM model,which is much faster,but it risks the patient’s life by showing 89.0% accuracy.Ensemble models consume more time than GBM because they consist of three models but show much higher accuracy results.The LDSVM with hard voting performs predictions in 0.59 s, 0.60 s with soft voting, and with the proposed approach, it shows a little bit faster by 0.55 s.

    Figure 7:Performance parameters result of different classifier

    Tab.2 display the results of Machine Learning and Proposed LDSVM with balancing technique along with a comparison of prediction times.With the suggested LDSVM,the shortest forecast time is 0.63 s,and the highest accuracy is reached.Fig.8 shows the result of using the PCA technique to reduce the feature count from 22285 columns to 48 new columns.

    A comparison of the proposed and previous techniques is discussed to illustrate the novelty of the proposed approach.Tab.3 presents machine learning and deep learning approaches proposed by different authors to classify leukemia.Machine learning models are less complex than deep learning models,which show faster training and prediction results than deep learning models.The accuracy of the first four related studies related to machine learning was higher than 95%.One of them achieved an accuracy rate of 99.1%.The last two studies were related to deep learning approaches,which showed the highest accuracy rate of 99.3%.

    Table 2:Comparative results of machine learning models and proposed LDSVM(Balancing)

    Figure 8:Cumulative variance against the number of components

    Therefore,the proposed study focuses on machine learning algorithms to find the best optimized,less complex,and fast prediction model that helps to classify leukemia in patients with accurate results to save patients’lives without false prediction.The proposed LDSVM is a combined model of LR,DT,and SVM by tuning and adjusting several hyperparameters that show the highest accuracy results of 99.8%.

    Table 3:Comparison of proposed LDSVM classifier with present approaches

    6 Conclusion

    The algorithms of machine learning are becoming more common.Leukemia cancer classification is proposed based on DNA microarray data using the proposed approach, which consists of a novel structure of classification methods.This study proposed machine-learning best algorithms for leukemia cancer classification.The three machine learning algorithms SVM, LR, and DT are combined using voting classifiers, named LDSVM and analyzing the classification results using these algorithms on leukemia.This study focuses on the voting classifier for accuracy, which works best on the chosen dataset.We have fulfilled our goal of accessing and investigating three selected algorithms based on the classification to predict the best leukemia pattern.The best leukemia-based voting classifier achieved the best accuracy.The primary objective of this study was to demonstrate the efficiency of voting classifiers in improving the classification accuracy.Three algorithms, SVM,LR, and DT, and their novel ensemble patterns were studied on the chosen dataset.Statistical measurements, such as accuracy, specificity, and sensitivity, were used to investigate the output of all the classifiers.The output of each classifier was also studied.Accuracy of the classification of three models and the ensemble model for training and testing data before and after the choice of functions.The proposed LDSVM model with hard voting achieves an accuracy of 0.95, and with a soft vote of 0.998,and with the balancing technique,an accuracy of 100%was achieved with minimum prediction time.The proposed approach achieves an accuracy result with 5 fold cross validation using grid search optimization for LDSVM to obtain outperform results.Voting classifier models work well in all respects, which enables our dataset to be categorized in this model.In future work, we propose that these observations suggest that the algorithm evaluated in machine learning can change the conventional classification methods for use in the medical field or,in general,bioinformatics.As the PCA technique works well for computation in the future,we need more data for better working of the model.

    Acknowledgement:The authors would like to thank the Editor and the anonymous reviewers for their insightful comments and constructive suggestions that certainly improved the quality of this paper.

    Funding Statement:The authors received no specific funding for this study.

    Conflicts of Interest:The authors declare that they have no conflicts of interest to report regarding the present study.

    亚洲精品中文字幕一二三四区| 欧美乱码精品一区二区三区| a级毛片在线看网站| 一本一本久久a久久精品综合妖精| 亚洲,欧美精品.| 亚洲精品美女久久久久99蜜臀| 男女床上黄色一级片免费看| 国产欧美亚洲国产| 免费高清在线观看日韩| 久久狼人影院| 天堂√8在线中文| 日本黄色日本黄色录像| 欧美日韩乱码在线| 午夜亚洲福利在线播放| 久久精品亚洲av国产电影网| 日本vs欧美在线观看视频| 99久久精品国产亚洲精品| 91麻豆av在线| 人人妻人人添人人爽欧美一区卜| 黑人操中国人逼视频| 欧美日韩亚洲国产一区二区在线观看 | 亚洲欧美一区二区三区久久| 国产精品99久久99久久久不卡| 亚洲成人免费电影在线观看| 久久久国产欧美日韩av| 免费在线观看完整版高清| 久久香蕉国产精品| 叶爱在线成人免费视频播放| 成在线人永久免费视频| 亚洲七黄色美女视频| 中国美女看黄片| 欧美亚洲 丝袜 人妻 在线| 99精品久久久久人妻精品| 日本vs欧美在线观看视频| 欧美中文综合在线视频| 窝窝影院91人妻| 久久 成人 亚洲| www日本在线高清视频| 欧美成人午夜精品| 久久久国产欧美日韩av| 19禁男女啪啪无遮挡网站| 51午夜福利影视在线观看| 亚洲五月婷婷丁香| 亚洲免费av在线视频| www.熟女人妻精品国产| 91在线观看av| 欧美成人午夜精品| 久久精品熟女亚洲av麻豆精品| www.999成人在线观看| 叶爱在线成人免费视频播放| tocl精华| 国产精品免费大片| 中文字幕av电影在线播放| www日本在线高清视频| 国产99白浆流出| 国产在线观看jvid| 在线天堂中文资源库| 亚洲国产精品一区二区三区在线| 一区在线观看完整版| 亚洲免费av在线视频| 最新美女视频免费是黄的| 99国产精品99久久久久| 成人特级黄色片久久久久久久| 亚洲午夜精品一区,二区,三区| 一级a爱视频在线免费观看| 欧美日韩亚洲高清精品| 亚洲精品一二三| videosex国产| 久久九九热精品免费| 性少妇av在线| 久久人人爽av亚洲精品天堂| 国产精品一区二区精品视频观看| 精品久久久久久,| 高清视频免费观看一区二区| 天天操日日干夜夜撸| 一区在线观看完整版| 熟女少妇亚洲综合色aaa.| 中文字幕av电影在线播放| 侵犯人妻中文字幕一二三四区| 好男人电影高清在线观看| 欧美黑人精品巨大| 久久中文字幕一级| 国产精品 国内视频| 男人操女人黄网站| 精品卡一卡二卡四卡免费| av天堂在线播放| 国产精品自产拍在线观看55亚洲 | 国产深夜福利视频在线观看| 久久人妻熟女aⅴ| 中文字幕色久视频| 一级黄色大片毛片| 亚洲成人国产一区在线观看| 别揉我奶头~嗯~啊~动态视频| 精品国产乱子伦一区二区三区| 不卡一级毛片| 国产亚洲欧美在线一区二区| 久久久国产成人精品二区 | 丁香欧美五月| 又大又爽又粗| 丝袜美足系列| 亚洲美女黄片视频| 中亚洲国语对白在线视频| av一本久久久久| 亚洲精品在线美女| 成年人免费黄色播放视频| 亚洲人成电影免费在线| 亚洲自偷自拍图片 自拍| 搡老岳熟女国产| 大型av网站在线播放| 国内久久婷婷六月综合欲色啪| 又紧又爽又黄一区二区| 久久香蕉国产精品| 亚洲成a人片在线一区二区| 久久精品亚洲精品国产色婷小说| 日韩欧美一区二区三区在线观看 | 岛国在线观看网站| 91精品三级在线观看| 女警被强在线播放| 国产免费男女视频| 欧美老熟妇乱子伦牲交| 成人黄色视频免费在线看| 窝窝影院91人妻| 日本五十路高清| 少妇裸体淫交视频免费看高清 | 国产精品一区二区免费欧美| 日韩免费av在线播放| 亚洲人成77777在线视频| 亚洲成a人片在线一区二区| 老司机亚洲免费影院| 午夜福利乱码中文字幕| 99热国产这里只有精品6| 精品卡一卡二卡四卡免费| 欧美精品高潮呻吟av久久| 免费人成视频x8x8入口观看| 看免费av毛片| 91成人精品电影| 久久99一区二区三区| 欧美国产精品一级二级三级| 精品电影一区二区在线| 大型av网站在线播放| 成人特级黄色片久久久久久久| 免费在线观看完整版高清| 女警被强在线播放| 超碰成人久久| 中文字幕人妻丝袜一区二区| 国产麻豆69| 欧美 亚洲 国产 日韩一| 日韩欧美一区视频在线观看| 亚洲人成电影免费在线| av免费在线观看网站| 人妻一区二区av| 国产极品粉嫩免费观看在线| www日本在线高清视频| 91麻豆精品激情在线观看国产 | 看片在线看免费视频| 精品亚洲成a人片在线观看| 亚洲av日韩在线播放| 高清av免费在线| 激情视频va一区二区三区| 岛国毛片在线播放| 久久久久久亚洲精品国产蜜桃av| 999久久久精品免费观看国产| 日韩熟女老妇一区二区性免费视频| 成年女人毛片免费观看观看9 | 久久久国产精品麻豆| 自拍欧美九色日韩亚洲蝌蚪91| 黑人巨大精品欧美一区二区mp4| 一二三四社区在线视频社区8| 免费在线观看黄色视频的| 欧美久久黑人一区二区| 精品电影一区二区在线| 在线观看午夜福利视频| 搡老熟女国产l中国老女人| 国产真人三级小视频在线观看| 91国产中文字幕| 看黄色毛片网站| 欧美在线黄色| 欧美日韩一级在线毛片| 水蜜桃什么品种好| 99国产综合亚洲精品| 国产精品免费大片| 新久久久久国产一级毛片| av网站免费在线观看视频| 久久人人爽av亚洲精品天堂| 国产不卡一卡二| 久久精品亚洲av国产电影网| 久久精品亚洲熟妇少妇任你| 亚洲精品美女久久av网站| 久久久久视频综合| 亚洲精品在线美女| 亚洲黑人精品在线| 热re99久久国产66热| 久久亚洲精品不卡| 日本wwww免费看| 国产精品99久久99久久久不卡| 成年人免费黄色播放视频| 久久久久视频综合| 999久久久国产精品视频| 99国产精品一区二区蜜桃av | 日韩三级视频一区二区三区| 久久久国产一区二区| 久久久久久免费高清国产稀缺| 国产精品一区二区在线观看99| 国产97色在线日韩免费| 亚洲精品av麻豆狂野| 亚洲熟女毛片儿| av中文乱码字幕在线| 亚洲中文av在线| 18禁裸乳无遮挡动漫免费视频| 香蕉久久夜色| 午夜影院日韩av| 久久精品国产亚洲av高清一级| 久久久国产成人精品二区 | 涩涩av久久男人的天堂| 成人手机av| 亚洲一区二区三区不卡视频| 捣出白浆h1v1| 国产成人啪精品午夜网站| 飞空精品影院首页| 免费人成视频x8x8入口观看| 一本综合久久免费| 黄片小视频在线播放| 免费不卡黄色视频| 动漫黄色视频在线观看| 黄频高清免费视频| 国精品久久久久久国模美| 国产一区有黄有色的免费视频| 大香蕉久久成人网| 一本大道久久a久久精品| 99re在线观看精品视频| 久久久国产成人免费| 国产精品乱码一区二三区的特点 | 国产真人三级小视频在线观看| 麻豆国产av国片精品| 久久精品亚洲熟妇少妇任你| 国产成人精品在线电影| 看免费av毛片| av视频免费观看在线观看| 色婷婷av一区二区三区视频| 国产精品综合久久久久久久免费 | 国产免费av片在线观看野外av| 国产成人av教育| 亚洲五月色婷婷综合| 午夜福利在线免费观看网站| 精品一区二区三区av网在线观看| av在线播放免费不卡| 三级毛片av免费| 欧美成人午夜精品| 免费人成视频x8x8入口观看| 一本大道久久a久久精品| 天堂√8在线中文| 国产精品欧美亚洲77777| 嫩草影视91久久| 亚洲一区二区三区欧美精品| 久久精品亚洲熟妇少妇任你| 91成年电影在线观看| 欧美精品av麻豆av| 国产不卡一卡二| 免费在线观看亚洲国产| 少妇粗大呻吟视频| 欧美成狂野欧美在线观看| 曰老女人黄片| 男女免费视频国产| 女人爽到高潮嗷嗷叫在线视频| 中文字幕人妻丝袜一区二区| 久久影院123| 日韩成人在线观看一区二区三区| 人人妻人人澡人人看| 精品少妇一区二区三区视频日本电影| 啪啪无遮挡十八禁网站| 757午夜福利合集在线观看| 久久精品人人爽人人爽视色| 精品国产乱子伦一区二区三区| 欧美黑人精品巨大| 国产亚洲欧美精品永久| 免费黄频网站在线观看国产| 国产成人一区二区三区免费视频网站| 老汉色∧v一级毛片| 免费人成视频x8x8入口观看| 女性被躁到高潮视频| 亚洲av电影在线进入| 欧美久久黑人一区二区| 亚洲精品一二三| 欧美另类亚洲清纯唯美| 国产一区在线观看成人免费| 狂野欧美激情性xxxx| ponron亚洲| 桃红色精品国产亚洲av| 美女国产高潮福利片在线看| 国产亚洲精品久久久久5区| 亚洲色图综合在线观看| 人人妻人人爽人人添夜夜欢视频| 亚洲精品国产色婷婷电影| 国产真人三级小视频在线观看| 免费在线观看亚洲国产| 久久精品国产a三级三级三级| 亚洲成a人片在线一区二区| 侵犯人妻中文字幕一二三四区| а√天堂www在线а√下载 | 欧美精品一区二区免费开放| 国产亚洲av高清不卡| 欧美在线一区亚洲| 亚洲午夜精品一区,二区,三区| 99国产综合亚洲精品| 国产成人免费无遮挡视频| 亚洲第一青青草原| 伦理电影免费视频| 久久精品国产清高在天天线| svipshipincom国产片| 日韩成人在线观看一区二区三区| 欧美丝袜亚洲另类 | 美女国产高潮福利片在线看| 亚洲av日韩在线播放| 国产亚洲精品一区二区www | 在线观看66精品国产| 黄频高清免费视频| 国产精品乱码一区二三区的特点 | 国产一区二区三区综合在线观看| 亚洲欧美日韩高清在线视频| 午夜福利,免费看| 法律面前人人平等表现在哪些方面| 无遮挡黄片免费观看| 国产国语露脸激情在线看| 欧美亚洲日本最大视频资源| 欧美精品一区二区免费开放| 不卡一级毛片| 久9热在线精品视频| 亚洲精品成人av观看孕妇| 高清黄色对白视频在线免费看| 美国免费a级毛片| 女性生殖器流出的白浆| 啦啦啦视频在线资源免费观看| 欧美激情 高清一区二区三区| 亚洲精品国产一区二区精华液| 国产精品久久久av美女十八| 国产成人欧美在线观看 | 伦理电影免费视频| 国产91精品成人一区二区三区| 国产精品欧美亚洲77777| av在线播放免费不卡| 大码成人一级视频| avwww免费| 中文字幕色久视频| 国产在线观看jvid| 国产成人欧美在线观看 | 亚洲精品国产一区二区精华液| 天天影视国产精品| 国产单亲对白刺激| 午夜福利欧美成人| 欧美国产精品一级二级三级| 国产主播在线观看一区二区| 黄色丝袜av网址大全| 日韩精品免费视频一区二区三区| 亚洲精品国产一区二区精华液| 亚洲第一av免费看| 色综合欧美亚洲国产小说| 天天影视国产精品| 午夜久久久在线观看| 国产xxxxx性猛交| 99久久综合精品五月天人人| 国产xxxxx性猛交| 777米奇影视久久| 精品久久久久久,| 国产成人欧美在线观看 | 最近最新中文字幕大全电影3 | 国产精品免费大片| 他把我摸到了高潮在线观看| 中文字幕最新亚洲高清| 男女床上黄色一级片免费看| 国产成人av激情在线播放| 亚洲第一青青草原| 亚洲 欧美一区二区三区| 人妻一区二区av| 久久久久久久久免费视频了| 久久精品国产亚洲av香蕉五月 | av有码第一页| 热re99久久国产66热| 午夜福利免费观看在线| 亚洲一区二区三区不卡视频| 久久99一区二区三区| 亚洲av成人一区二区三| 欧美日韩亚洲综合一区二区三区_| 美国免费a级毛片| 狠狠婷婷综合久久久久久88av| 美国免费a级毛片| 99国产综合亚洲精品| 免费观看精品视频网站| 91成年电影在线观看| 黑人操中国人逼视频| 欧美日韩亚洲综合一区二区三区_| 国产成+人综合+亚洲专区| 久久中文字幕一级| 国产精品国产高清国产av | 国产精品综合久久久久久久免费 | 久久国产精品大桥未久av| 午夜日韩欧美国产| 国产伦人伦偷精品视频| 色老头精品视频在线观看| 电影成人av| 亚洲欧美精品综合一区二区三区| 久久性视频一级片| 亚洲专区中文字幕在线| 黄色片一级片一级黄色片| 人人妻,人人澡人人爽秒播| 亚洲人成伊人成综合网2020| 午夜福利视频在线观看免费| 国产精品久久电影中文字幕 | 久久久国产成人精品二区 | 母亲3免费完整高清在线观看| 久热爱精品视频在线9| 国产黄色免费在线视频| 国产99白浆流出| 在线观看日韩欧美| 男女午夜视频在线观看| 国产精品免费大片| 欧美黄色片欧美黄色片| 在线观看免费高清a一片| 亚洲性夜色夜夜综合| 老司机深夜福利视频在线观看| 天堂中文最新版在线下载| 欧美精品啪啪一区二区三区| av超薄肉色丝袜交足视频| 国产亚洲精品久久久久久毛片 | tube8黄色片| 777米奇影视久久| 黄色女人牲交| 一级毛片女人18水好多| 91成年电影在线观看| 久久精品亚洲精品国产色婷小说| 国产精品 国内视频| 日本黄色视频三级网站网址 | 在线观看一区二区三区激情| 日本欧美视频一区| 精品一区二区三区四区五区乱码| 国产精品久久久av美女十八| √禁漫天堂资源中文www| 丰满的人妻完整版| 国产精品电影一区二区三区 | 欧美不卡视频在线免费观看 | 精品久久久久久久久久免费视频 | 人成视频在线观看免费观看| 激情视频va一区二区三区| 亚洲色图 男人天堂 中文字幕| 男人舔女人的私密视频| 欧美黄色淫秽网站| 亚洲精品久久成人aⅴ小说| 久久热在线av| 成人亚洲精品一区在线观看| 欧美精品啪啪一区二区三区| 最近最新中文字幕大全免费视频| 美国免费a级毛片| 99国产综合亚洲精品| netflix在线观看网站| 国产男女超爽视频在线观看| 99精品在免费线老司机午夜| 一本大道久久a久久精品| 国产精品.久久久| 夜夜躁狠狠躁天天躁| 国产精品香港三级国产av潘金莲| 在线av久久热| 99久久人妻综合| 制服诱惑二区| 国产亚洲精品久久久久久毛片 | 老熟妇乱子伦视频在线观看| 男女之事视频高清在线观看| 久久精品国产综合久久久| 久久精品人人爽人人爽视色| 9色porny在线观看| 国产av一区二区精品久久| 久久久国产一区二区| 亚洲第一av免费看| 久久久久久亚洲精品国产蜜桃av| av线在线观看网站| 国产国语露脸激情在线看| 欧美日韩亚洲高清精品| 欧美精品一区二区免费开放| 动漫黄色视频在线观看| 久久精品成人免费网站| 亚洲专区国产一区二区| 一级,二级,三级黄色视频| 国产成人欧美| 国产区一区二久久| 亚洲视频免费观看视频| 丝袜美足系列| 亚洲五月天丁香| 黄网站色视频无遮挡免费观看| bbb黄色大片| tube8黄色片| 成人影院久久| 视频区欧美日本亚洲| 亚洲国产精品合色在线| 亚洲熟女精品中文字幕| 男女午夜视频在线观看| 久久ye,这里只有精品| 人人妻人人爽人人添夜夜欢视频| 精品一区二区三区av网在线观看| 男女床上黄色一级片免费看| 久久人妻福利社区极品人妻图片| 日韩欧美在线二视频 | 一级片免费观看大全| 精品福利观看| 国产有黄有色有爽视频| 老司机亚洲免费影院| 男女下面插进去视频免费观看| 色老头精品视频在线观看| 亚洲七黄色美女视频| 亚洲欧美一区二区三区久久| 亚洲中文av在线| 啦啦啦视频在线资源免费观看| 亚洲成av片中文字幕在线观看| 淫妇啪啪啪对白视频| 热99久久久久精品小说推荐| 久久久久精品国产欧美久久久| 在线永久观看黄色视频| 多毛熟女@视频| 国产欧美日韩一区二区三| 伦理电影免费视频| 午夜免费观看网址| av天堂久久9| 国产一卡二卡三卡精品| 国产精品一区二区在线不卡| 国产不卡av网站在线观看| 成人18禁高潮啪啪吃奶动态图| 午夜福利,免费看| 人妻丰满熟妇av一区二区三区 | 极品人妻少妇av视频| 亚洲片人在线观看| 美女福利国产在线| 麻豆av在线久日| 丝瓜视频免费看黄片| 亚洲精品国产一区二区精华液| 啦啦啦 在线观看视频| 不卡av一区二区三区| 国产精品成人在线| 国产成人系列免费观看| 免费观看精品视频网站| 99国产精品一区二区蜜桃av | 久久精品国产99精品国产亚洲性色 | 亚洲欧美日韩高清在线视频| 亚洲av第一区精品v没综合| 999精品在线视频| 人妻久久中文字幕网| 国产欧美日韩精品亚洲av| 欧美日韩瑟瑟在线播放| 99热国产这里只有精品6| 国产主播在线观看一区二区| 精品高清国产在线一区| 18在线观看网站| 午夜影院日韩av| 久久久久久久久免费视频了| а√天堂www在线а√下载 | 精品午夜福利视频在线观看一区| 成年人免费黄色播放视频| 曰老女人黄片| 日本wwww免费看| 亚洲精品一二三| 久久精品亚洲精品国产色婷小说| 制服诱惑二区| 国产淫语在线视频| 热re99久久国产66热| 这个男人来自地球电影免费观看| 国产野战对白在线观看| 久久人妻av系列| 性色av乱码一区二区三区2| 飞空精品影院首页| 夫妻午夜视频| 99精国产麻豆久久婷婷| 国产99白浆流出| 人妻久久中文字幕网| 视频区欧美日本亚洲| 国产亚洲av高清不卡| 亚洲精品乱久久久久久| 中文字幕人妻丝袜一区二区| 激情在线观看视频在线高清 | 日韩欧美一区视频在线观看| 亚洲午夜理论影院| 99riav亚洲国产免费| 国产一区有黄有色的免费视频| 精品一区二区三区av网在线观看| 男女下面插进去视频免费观看| 男人舔女人的私密视频| 国产精品一区二区在线不卡| 性色av乱码一区二区三区2| 一进一出抽搐gif免费好疼 | 亚洲精品成人av观看孕妇| 国产精品亚洲av一区麻豆| 亚洲精品美女久久av网站| av免费在线观看网站| 夜夜夜夜夜久久久久| 久久久久精品人妻al黑| 成年版毛片免费区| 女警被强在线播放| 久久中文看片网| 欧美激情久久久久久爽电影 | 亚洲一卡2卡3卡4卡5卡精品中文| 国产高清激情床上av| 精品久久久精品久久久| 69精品国产乱码久久久| 国产蜜桃级精品一区二区三区 | 一区二区三区精品91| 99re在线观看精品视频| 亚洲av成人一区二区三| 如日韩欧美国产精品一区二区三区| 国产真人三级小视频在线观看| 中文字幕高清在线视频| 两性夫妻黄色片| 日韩欧美一区视频在线观看| 日韩免费高清中文字幕av| 精品熟女少妇八av免费久了| 国产一卡二卡三卡精品| 又黄又爽又免费观看的视频| 日本撒尿小便嘘嘘汇集6| 99精品欧美一区二区三区四区| 又黄又爽又免费观看的视频| aaaaa片日本免费|