48 results about "Information gain ratio" patented technology

The invention relates to an unbalanced data sampling method in an improved C4.5 decision tree algorithm. The method comprises the steps as follows: firstly, initial weights of various samples are determined according to the number of various samples; the weights of the samples are modified through the training result of the improved C4.5 decision tree algorithm in each round; the information gain ratio and misclassified sample weights are taken into account by a division standard of the improved C4.5 algorithm; the final weights of the samples are obtained after T iterations; the samples in minority class boundary regions and majority class center regions are found out according to the sample weights; over-sampling is carried out on the samples in the minority class boundary regions by an SMOTE algorithm; and under-sampling is carried out on majority class samples by a weight sampling method, so that the samples in the center regions are relatively easily selected to improve the balance degree of different classes of data, and the recognition rates of the minority class and the overall data set are improved. According to the unbalanced data sampling method in the improved C4.5 decision tree algorithm, weight modification is carried out through the improved C4.5 decision tree algorithm; and over-sampling and under-sampling are specifically carried out according to the sample weights, so that the phenomena of classifier over-fitting, loss of useful information of the majority class and the like are effectively avoided.

The invention discloses an advanced persistent threat APT attack detection method. A semi-supervised learning algorithm is used to mark data having similar characteristics, and a small quantity of marked data are used to generate a large-scale training data set, and then an information gain ratio is introduced to determine the degrees of the influences of the different characteristics on detection. The information gate ratio is used for characteristic extraction of every sub-data set in a detection model, and accurate identification of unknown attacks is realized. An improved k-means algorithm is used to mark the data having the similar characteristics, and on the basis of the small quantity of marked data, the accurate marking of a lot of training data sets is realized, and therefore the detection accuracy of the model is guaranteed; and by introducing the information gain ratio in the model, the degrees of the influences of the different characteristics on the detection are determined, and influences of redundancy and noise characteristics in the data are reduced, and therefore important flow characteristics are selected, the generalization capability of the detection model is improved to detect the unknown attacks.

The invention discloses an integrated learning method for recognizing ECM (extracellular matrix) protein. According to the method, data set building: a training sample set and an independent test sample set of an ECM protein sequence are built; the protein sequence in the training sample set is mapped into a numeric feature vector; a relatively effective feather subset is selected by an information gain ratio-incremental feature selection method, an integrated classifier model is built by an integrated learning method, and the problem of data set unbalance is solved; the independent test sample set is mapped into a numeric feature vector, the category of the test sample is obtained by a majority voting method on the basis of a predication result of the integrated learning method, and the performance of a prediction system is finally evaluated by utilizing the predication result of the test sample. The invention discloses a network server system for recognizing the ECM protein. Users do not need to understand the concrete executing process of ECM protein recognition, and the prediction result can be obtained only through inputting the protein sequence to be predicted.

The invention discloses a detection method for automatically identifying syndrome types of traditional Chinese medicine, which comprises the following steps: establishing a standard and objective database of TCM cases; aiming at the standardized TCM sample database, using an attribute screening method based on association relations to calculate the mutual relationship between each attribute Information and symmetric uncertainty, based on heuristic rules, select the symptom attribute set that contributes more to syndrome detection; use the selected key attribute set and sample information in the case database to construct a classification training sample set, and calculate the attribute information gain rate, determine the decision-making attributes, control the sample lower limit of each node and record the classification error at the same time, read all training samples and quasi-training samples in the way of incremental learning, and finally obtain the classification rules; use the obtained classification rules to carry out new Sample identification test. The invention can not only be applied to the problem of automatic syndrome differentiation of liver cirrhosis, but also can be extended to the field of automatic discrimination of other syndrome types in traditional Chinese medicine.

The embodiment of the invention discloses a user behavior prediction method and device and a storage medium. The method comprises the steps of obtaining attribute data of an appointed attribute of a target user; inputting the attribute data of the specified attribute into the target behavior prediction model to obtain a target behavior prediction result of the target user. The target behavior prediction model is obtained through the following steps: obtaining user data of a sample set, wherein the user data comprises to-be-screened user attributes and user target behavior historical data; screening out a plurality of user attributes based on the information gain and the information gain ratio of the user attributes; determining a possibility estimated value of a target behavior of the userunder each user attribute, and further screening out a plurality of user attributes to form a specified attribute set; and training a preset machine learning model according to the user data corresponding to the specified attribute of the sample set. According to the embodiment of the invention, the accuracy of user behavior prediction can be improved.

The invention discloses a diabetes mellitus probability calculation method based on large data of a diabetes mellitus system. The diabetes mellitus probability calculation method comprises the following steps: (1) constructing a diabetes mellitus decision tree model; (2) selecting an optimal branch variable of the decision tree model according to the information gain ratio Gains (Xi) of a training sample S; (3) post-pruning a decision tree from bottom to top; and (4) constructing a diabetes mellitus naive Bayes model, and obtaining the diabetes mellitus probability P (C1|y1.y2.y3.....ym) with C1 serving as an output variable on a rth node by utilizing a Bayes formula. According to the diabetes mellitus probability calculation method disclosed by the invention, a two-layer model method with the combination of the decision tree and the naive Bayes model is designed; by extracting the characteristic attribute of diabetes mellitus from the large data, whether the diabetes mellitus occurs or not is forecasted, the probability of the occurrence of the diabetes mellitus is further calculated, and the prevention and the forecast are combined and are relatively overall and accurate.

The embodiment of the invention discloses a cyclotella identification method and apparatus. The method comprises the steps of performing slice processing on a high-magnification microscopic image of an original water sample containing a to-be-tested alga to obtain a square identification image with a cyclotella shape as an internally tangent circle; performing polar coordinate transformation on the identification image, and extracting multiple identification characteristics of the image after the polar coordinate transformation; and inputting the extracted multiple identification characteristics to a decision-making tree model constructed in advance, and determining a category of the to-be-tested alga according to a decision-making tree model output result, wherein the decision-making treemodel performs classification using a C4.5 decision tree algorithm according to characteristics of a template cyclotella image and an information gain ratio of each training sample characteristics ina training sample image set, and performs identification on the to-be-tested alga according to a classification result. According to the technical scheme of the invention, that incision cannot be formed on an alga image under a complex background according to the prior art for identification is effectively solved, and the identification accuracy rate and efficiency on the alga are improved.

The invention discloses a distributed denial of service attack detection method based on a C4.5 decision tree algorithm in software defined network environment, and the method comprises the followingsteps: collecting flow table information returned back by an OpenFlow switch through an OpenFlow protocol; extracting field information related to a DDoS attack from the flow table information, converting the extracted information into parameters capable of analyzing network flow distribution variation and taking the parameters as attributes, and forming a training set of a decision tree; classifying flows with the C4.5 decision tree algorithm, calculating class information entropy according to training set data classes; orderly calculating conditional entropy of the attributes, gain of information, information entropy of the attributes and information gain ratio of the attributes; selecting the attribute with the highest information gain ratio as a root node of the decision tree, and selecting the attributes with highest information gain ratio from the residual attributes as a fork node, and repeating the steps above until forming the decision tree; and using the finally formed decision tree to perform classification operation for the new network flow, and detecting whether the DDoS attack exists. The method can detect the DDoS attack more accurately.

The invention provides a method and system for acquiring the importance levels of evaluation indexes. The method comprises the following steps of: saving evaluation results of n evaluation indexes X1, X2, ellipsis, Xi, ellipsis, and Xn obtained by using m systems respectively in a data set S, saving the risk level of each system in a data set R, and calculating entropy H (R) of the risk level in the data set R; and acquiring the information gain ratios of n evaluation indexes, and sequencing the acquired information gain ratio to obtain the importance levels of n evaluation indexes X1, X2, ellipsis, Xi, ellipsis, and Xn.

The invention provides a structure extended polynomial naive Bayes text classification method. Firstly, a one-dependence polynomial estimator is established by using each word that occurs in a test document as a father node and then all the one-dependence polynomial estimators are subjected to weighted averaging to predict a category of the test document, wherein the weight is an information gain ratio of each word. According to the method, the structure learning phase of a Bayesian network is avoided, thereby reducing time spending brought by high dimensionality of text data; and meanwhile, the estimation process of a dual conditional probability is postponed to the classification stage, thereby ingeniously saving large space cost. According to the method, not only is classification accuracy of a polynomial naive Bayes text classifier improved, but also time spending and space cost of structure learning of the Bayesian network are avoided.

The invention discloses a case reasoning classifier case retrieval method. The method comprises the steps of S1, case library protocol feature selection; S2, clustering decision tree training on a C4.5 algorithm and a case library; and S3, a weighted voting KNN method. According to the method, improved research is carried out on a case retrieval method of a case reasoning classifier: a method of integrating and using a inductive index strategy in neighbor case retrieval is provided, and a feature evaluation method of an information gain ratio is adopted to select a feature training case library clustering decision tree with a relatively remarkable information gain ratio; during case retrieval, a cluster corresponding to a target case is retrieved according to the clustering decision tree, and then neighbor case retrieval is performed in the case cluster. Experimental data of the open source data set show that compared with a traditional CBR model, the improved model can fully utilize the classification capability of case characteristics to carry out case retrieval, the calculation time complexity of the CBR reasoning classifier is effectively reduced, and the reasoning accuracy is improved.

The invention provides a rock bust danger level prediction method based on a local weighting C4.5 algorithm and relates to the technical field of rock burst prediction. The method includes the steps of firstly, adopting an MDLP method for conducting discretization on continuous attribute data in sample data, then adopting a local weighting method for selecting a training set and calculating the weight of samples, utilizing the weight of the samples to calculate an information gain ratio of each attribute, and selecting sample attributes as root nodes of a C4.5 decision tree and splitting attributes of other branch nodes according to the information gain ratios; finally, adopting the weight of the samples to substitute the sample number to conduct pessimistic pruning on the created decisiontree, and correspondingly achieving prediction of rock burst dangers and the like in a predicted area. According to the provided rock bust danger level prediction method based on the local weightingC4.5 algorithm, the defect is overcome that the preference selection values have too many attributes when information gain is adopted for selecting node splitting attributes in an ID3 algorithm; an over-fitting problem is avoided, and the prediction accuracy of a model is high.

The invention discloses a hidden Markov chain model based intelligent recommendation algorithm, which is applied to an intelligent recommendation system of a law net. The algorithm comprises the steps of for characteristics of documents, replacing an initial probability and a state transfer probability of a state in a hidden Markov chain model with an information gain ratio and a correlation degree respectively; calculating a partial probability of reading a <t>th document by a client; calculating an optimal probability and an optimal document sequence of each document when a document number t is equal to n, and selecting the optimal probability with the maximum value from all the optimal probabilities; and recording the optimal document sequences, namely, all the document sequences recommended to the client. According to the implementation scheme provided by the invention, the advantages of the information gain ratio and a hidden Markov chain are combined, and the shortcoming of high recommendation deviation quantity due to single use of a hidden Markov chain algorithm as a recommendation algorithm and the shortcoming of complexity in parameter calculation are made up for, so that the hidden Markov chain model based intelligent recommendation algorithm has relatively high practicality and relatively high accuracy in recommendation algorithms.

The invention discloses an electric power business collaborative classification method and system based on a C4.5 decision tree algorithm, and the method comprises the following steps: obtaining an electric power business collaborative related database, and extracting a sample set S from the electric power business collaborative related database; extracting an index set A, wherein the index set Acontains indexes for evaluating the business collaboration data; calculating the information entropy and the information gain ratio of each index for the sample set S based on a C4.5 algorithm so as to select a proper root node and a proper intermediate node; constructing a decision tree according to the selected root node; evaluating and selecting each service cooperation scheme based on the decision tree. The invention further discloses a corresponding system. According to the invention, the information entropy and the information gain ratio are used for calculation, and classification rulesare easy to understand and high in accuracy. The method is applied to collaborative data calculation and analysis of businesses such as power outsourcing, optimal division features are selected as nodes to generate a decision-making tree and perform data classification, classification is quick and good in effect, and collaborative management of the businesses such as power outsourcing is effectively achieved.

The invention discloses an optimizing method and system of mobile service resources of an operator. The method comprises the steps that historical dialing data of a client of the operator are counted, wherein the dialing data are continuous variables; the continuous variables are converted into discrete characteristic variables through chi-square analysis; whether the client opens the mobile service or not is used as a classified variable of the two value, and a C4.5 decision-making tree model between the characteristic variables and the classified variable is established, wherein in the decision-making tree model, the information gain ratio corresponding to each kind of cutting is calculated and the cutting threshold with the maximum information gain ratio is selected as the optimal cutting threshold of the attribute; according to the decision-making tree model, values of the classified variable are calculated so as to obtain the prediction result of whether the client opens the mobile service or not; according to the prediction result, optimizing processing operation is conducted on the mobile service resources of the operator according to the prediction result. Through the technical scheme, the requirements for the mobile service of the client can be obtained efficiently according to dialing behaviors of the current client so as to realize optimized deployment of the mobile service resources of the operator.

The invention discloses a network business flow feature selecting and classifying method based on a multi-objective adaptive evolutionary algorithm, comprising the following steps of: firstly sortingthe features by using an information gain ratio and filtering part of irrelevant features to rapidly reduce dimension, then searching a feature space according to the adaptive evolutionary algorithm,using the feature having the top-raking information gain ratio as an initial population, and regarding two target functions of an inconsistent rate and a feature subset dimension as evaluation functions for selecting an optimal feature subset. Adaptive crossover and mutation maintain the population diversity, and the convergence ability of the algorithm is ensured. At the same time, by using a designed three-layer KNN classifier model, the invention classifies six multimedia business flows of online standard-definition live video, web browsing (Baidu), online audio, web browsing (sina), network voice chat and online standard-definition non-live video. The experiment result shows that the invention has higher classification accuracy compared to existing method.

The invention discloses a mutual information-based data discretization and feature selection integrated method. The method comprises the steps of 1) generating a proper candidate breakpoint set by performing breakpoint analysis on data; 2) searching for an optimal breakpoint subset in the candidate breakpoint set by adopting forward search, and evaluating the breakpoint subset by calculating mutual information between a data division result and original tag distribution of the data; 3) defining a search stop condition that the information gain ratio is less than a preset threshold or the total number of selected breakpoints exceeds the preset threshold; and 4) performing discretization and feature selection on the data by using the optimal breakpoint subset. The invention furthermore discloses a mutual information-based data discretization and feature selection integrated apparatus. According to the mutual information-based data discretization and feature selection integrated method and apparatus, the data discretization and feature selection processes are organically integrated, so that unrelated and redundant information in the data can be effectively removed and the performance of a subsequent learning algorithm can be improved.

The invention discloses a customer risk feature screening method based on SVMRFE and application thereof. The method comprises the following steps: acquiring customer risk feature data comprising a plurality of customer features; for the customer risk feature Xj, calculating an importance comprehensive measurement index of the customer risk feature Xj based on the Gini coefficient, the informationgain, the information gain ratio, the mutual information and the feature weight of the optimal classification result of the SVM classifier; after importance comprehensive measurement indexes of all customer risk characteristics are calculated in sequence, sorting is carried out in sequence from large to small according to the indexes, and the first k characteristics are selected to form a customer risk characteristic set. According to the method, the correlation between the selected feature subset and the target variable and the redundancy of the feature subset are considered, the defect thatthe customer risk features are screened by using a single index is overcome, and the training result of the SVM classifier is introduced in the feature screening process, so that the screened customer risk features are more suitable for the characteristics of the SVM classifier, and application prospects are good.

The invention relates to the technical field of data storage, and discloses a cloud environment data storage optimization method based on an HDFS, which comprises the following steps: putting metadataand storage data of each node of an HDFS cluster into a pre-created data volume container; calculatingavailable storage space evaluation values of the HDFS cluster nodes in the physical machines andthe data volume containers respectively; calculating an availability value of each physical machine, and calculating a performance evaluation value of the HDFS cluster node based on the physical volume container according to the availability value of each physical machine; storing the data blocks by using a data storage copy placement algorithm; calculating a data attribute information gain ratioof the to-be-stored data; dividing to-be-stored data blocks in the HDFS cluster by utilizing a KADC-KNN algorithm based on an information gain ratio and weighting; and for the to-be-stored data blocksdivided by the KADC-KNN algorithm based on the information gain ratio and the weighting, storing the to-be-stored data blocks into a Feeder HDFS cluster according to different storage strategies. According to the invention, the optimization of data storage is realized.

The invention provides a feature selection method based on an information gain ratio. The method comprises the steps of sorting attributes according to the size of information gain ratio of each attribute; determining the number of selection attributes by implementing a 5-fold cross validation method for 9 times, i.e. percentage; and finally building a Naive Bayes text classifier on a selected attribute subset. According to the feature selection method based on the information gain ratio provided by the invention, the method is a mixed attribute selection method by integrating the advantages of a filtering method and a packaging method; furthermore, experimental results of a plurality of standard text classification datasets show that the classification precisions of the Naive Bayes text classifier can be improved by the feature selection method based on the information gain ratio in most situations; meanwhile, too much time expenses cannot be caused.

The invention discloses a mobile phone feature detection optimization improvement algorithm based on a C4.5 decision tree. The method is based on a C4.5 algorithm, and in the process of calculating the information gain and the information gain ratio for each factor, a weight parameter is given aiming at the division of different factors, so that the information gain degree data difference of different factors is greater than that of the prior art, the recognition of machine learning on the use characteristic difference of different users is improved, and the credibility of a result is improved. According to the method and the device, the use characteristics of the mobile phone user are analyzed and recorded, and the characteristics of the current mobile phone user are matched with the characteristics of the mobile phone user, so that whether the mobile phone user is legal or not is judged, and the mobile phone cannot be illegally used by a third-party user. The invention aims to improve the mobile phone feature detection accuracy by improving and optimizing a C4.5 decision tree machine learning algorithm and more accurately identify illegal use operations of the mobile phone.

The invention discloses a method for determining the optimal segmentation scale of satellite image segmentation based on information gain rate, includes obtaining high-resolution satellite remote sensing image, preprocessing, selecting representative region as sample region determined by optimal segmentation scale, classifying surface features of sample region, obtaining image classification result of sample region as reference image determined by optimal segmentation scale, and obtaining image classification result of sample region as reference image determined by optimal segmentation scale;setting a series of segmentation scale parameters from small to large, using multi-scale segmentation technology, the sample image is segmented in multi-scale to obtain a series of segmented object vectors; Based on Shannon information entropy formula, the information entropy of reference image and segmentation vector is calculated, and the conditional entropy of reference image with segmentationvector is calculated, then the information gain and information gain ratio of segmentation vector are calculated. The optimal segmentation scale is selected according to the principle of maximum information gain rate. Finally, based on the optimal segmentation scale, the original image is segmented at multiple scales.

The invention provides a method and a system for configuring a business expansion metering device based on a knowledge graph. The method comprises the following steps of: S1, acquiring business expansion information of historical users and operation data corresponding to the business expansion information; S2, constructing a typical configuration information table according to the business expansion information; S3, calculating the information gain ratio of each type of information in the configuration information table, and constructing a metering device configuration decision tree; S4, obtaining a business expansion demand of a new user, and generating a plurality of configuration schemes corresponding to the business expansion demand through the metering device configuration decision tree; and S5, performing quantitative evaluation on the configuration scheme according to the operation data, generating an evaluation report, and performing visual display. By means of the mode, the configuration scheme obtained by the user can better meet the actual requirement while the configuration scheme making efficiency is improved.

In accordance with an embodiment, described herein is a system and method for automatic inference of a cube schema from a tabular data for use in a multidimensional database environment. A cube schemainference component can successively perform column-wise splits to calculate information gain ratios between each pair of a plurality of columns in the tabular data. A cross correlation matrix can beconstructed from the information gain ratios between each pair of columns. The system can determine relationships among the plurality of columns based on the cross correlation matrix; create a hierarchy directed graph to represent the relationships; and further transform the hierarchy directed graph into a cube schema that can be used to create a cube for loading the tabular data, or to map the tabular data into an existing cube in the multidimensional database environment.

The invention belongs to the field of security data analysis, and particularly relates to a feature selection method based on fuzzy set feature entropy calculation. The method mainly comprises the steps of calculating an ideal vector matrix, calculating a similarity matrix, calculating entropy of features, calculating a scaling factor SFi, and calculating a scaling factor SFi. According to the method, the distance between the ideal vectors in each category is calculated by using the scaling factors of the features and the entropy between the specific categories, feature selection can be optimized, and the calculation complexity is reduced. According to the method, a fuzzy set information entropy calculation method FIEE is adopted to solve the problem that in a traditional information gainand information gain ratio calculation method, calculation cannot be conducted due to the fact that the feature value space is huge. According to the method, the calculation complexity can be greatlyreduced.

The invention discloses a high-voltage circuit breaker fault diagnosis method based on a decision-making tree algorithm, and the method comprises the steps: feature selection: selecting feature vectors with classification capability for training data, wherein the feature vectors comprise four groups of feature vectors: a main contact stroke, a switching-on overtravel, action time and switching-onand switching-off synchronism; generating a decision tree, calculating an information gain ratio of the feature vectors in the training data set, judging whether the information gain ratio is maximumor not, selecting the feature vector with the maximum information gain ratio as a test condition of a decision node, and generating a complete decision tree according to a C4.5 algorithm; pruning thedecision-making tree, and pruning the decision-making tree according to the loss function of the minimized decision-making tree. Maintenance can be arranged in a targeted mode, unnecessary maintenancecan be effectively avoided, manpower and material resources are saved, and meanwhile power grid accidents caused by the fact that latent faults continue to be developed into dominant faults can be prevented.

The invention discloses a multi-feature fuzzy mapping access point optimization method based on an information gain ratio. The method comprises the following steps of: firstly, constructing an information gain ratio set of an access point (AP) and a fuzzy relation matrix of offline RSS characteristics by utilizing the offline RSS characteristics; secondly, obtaining fuzzy weights of the offline RSS characteristics by utilizing a fuzzy relation equation about the offline RSS characteristics; thirdly, acquiring RSS from different APs at a to-be-positioned point (namely a test point), and constructing a fuzzy judgment matrix of online RSS characteristics and a fuzzy membership set of APs in an online stage; and finally, defining the AP with a relatively large fuzzy membership degree as an APwith relatively strong position resolution, and using the AP as an optimized AP for positioning. Experimental results show that the method provided by the invention has relatively high positioning precision and relatively low calculation overhead.

The invention provides an information providing method and device. The method comprises the following steps: targeting to a first set, judging whether the number of each data item in the first set isgreater than a first threshold value or not; if so, calculating an information gain ratio of each selected filling parameter in the selected filling parameter list, selecting an inquiry template corresponding to the selected filling parameter with the maximum information gain ratio to inquire the user; selecting and filling parameter information; deleting the data items which are not matched withthe filling parameter information in the first set; obtaining a new first set, and when the data items in the new first set are not greater than the first threshold value, deleting the selected filling parameter with the maximum information gain ratio in the selected filling parameter list to obtain a new selected filling parameter list, re-judging whether the number of the data items in the new first set is greater than the first threshold value or not until the number of the data items is not greater than the first threshold value, and providing the data items in the new first set for the user. According to the method provided by the invention, the filling parameter with the maximum information gain ratio is selected each time to inquire the user, so that the frequency of interaction with the user is reduced, and the service efficiency is improved.

The invention discloses a distributed denial of service attack detection method based on a C4.5 decision tree algorithm in software defined network environment, and the method comprises the followingsteps: collecting flow table information returned back by an OpenFlow switch through an OpenFlow protocol; extracting field information related to a DDoS attack from the flow table information, converting the extracted information into parameters capable of analyzing network flow distribution variation and taking the parameters as attributes, and forming a training set of a decision tree; classifying flows with the C4.5 decision tree algorithm, calculating class information entropy according to training set data classes; orderly calculating conditional entropy of the attributes, gain of information, information entropy of the attributes and information gain ratio of the attributes; selecting the attribute with the highest information gain ratio as a root node of the decision tree, and selecting the attributes with highest information gain ratio from the residual attributes as a fork node, and repeating the steps above until forming the decision tree; and using the finally formed decision tree to perform classification operation for the new network flow, and detecting whether the DDoS attack exists. The method can detect the DDoS attack more accurately.