3264 results about "Random forest" patented technology

The invention discloses a fundus image retinal vessel segmentation method and a fundus image retinal vessel segmentation system based on deep learning. The fundus image retinal vessel segmentation method comprises the steps of performing data amplification on a training set, enhancing an image, training a convolutional neural network by using the training set, segmenting the image by using a convolutional neural network segmentation model to obtain a segmentation result, training a random forest classifier by using features of the convolutional neural network, extracting a last layer of convolutional layer output from the convolutional neural network, using the convolutional layer output as input of the random forest classifier for pixel classification to obtain another segmentation result, and fusing the two segmentation results to obtain a final segmentation image. Compared with the traditional vessel segmentation method, the fundus image retinal vessel segmentation method uses the deep convolutional neural network for feature extraction, the extracted features are more sufficient, and the segmentation precision and efficiency are higher.

The invention provides a method and a system for predicting cardiovascular and cerebrovascular disease risks. The method comprises the following steps: step 1, defining problems of cardiovascular and cerebrovascular disease prognosis risk prediction, step 2, collecting health medical data of cardiovascular and cerebrovascular disease patients; step 3, preprocessing data, including data integration, data cleaning, and processing missing data; step 4, constructing features and selecting features, identifying potential risk factors; step 5, the identified risk factors and rehabilitation outcomes forming an input-output sample set, inputting the input-output sample set to a random forest algorithm for model training, and evaluating prediction performance of a model. The method and the system can obtain health medical data of cardiovascular and cerebrovascular disease patients, the data being required by a clinician input model, and a predicted rehabilitation outcome of a patient in a certain time period in future is obtained through the model. The method and the system can preferably predict prognosis risks, so as to realize personalized accurate rehabilitation treatment.

The invention discloses a botnet detection method based on DNS (Domain Name System) flow characteristics. The method of the invention comprises a Domain-Flux botnet detection method based on DNS flow characteristics. The method comprises steps: legal main domain names and illegal main domain names are combined to form a target set; a domain name whose length is larger than 6 is extracted and processed as a research target; the domain name entropy, word formation characteristics, phonetic characteristics and grouped characteristics are calculated respectively; and the above is put in a random forest classifier to obtain a training model. A Fast-Flux botnet detection method based on the Domain-Flux botnet detection method comprises steps: the original data of a DNS server are processed; the training model obtained before is used for evaluating a to-be-processed domain name, and a score for a DGA condition is acquired; a white list, a black list and a grey list are used for scoring the domain name and an IP; time characteristics of the IP address are calculated; the stability of the IP address is calculated; and the above is put in the random forest classifier to obtain a training model SFF. The experiment accuracy is high.

The invention discloses a DGA domain name detection method based on random forest, comprising steps of constructing a knowledge database which comprises a black-and-white list sample library and a word dictionary, setting a domain name characteristic template, using the domain name in the black-and-white list as a training set, filtering the noise, performing training and off-line storage on the random forest algorithm model, obtaining a domain name to be detected, loading an optimal random forest algorithm model, and using the domain name to be detected as an input to obtain the prediction result. The invention does not rely on the DNS data which is obtained on line, can not only fast finish the DGA domain name detection, but also provides prediction to the other malicious domain name detection methods. Besides, the DGA domain name detection method is based on the forest algorithm and has an obvious advantage on the noise interference, and is less in resource consumption, high in operation efficiency and good in generalization.

The invention relates to an optimized classification method and an optimized classification device based on a random forest algorithm. The optimized classification method comprises the following steps of a first step, dividing given sample data into k sub-training sets which are independent from one another, selecting different decision tress according to each training sub-set, selecting different decision attributes by the decision trees for forming base classifiers, and forming a random forest by the base classifiers; a second step, in each base classifier, distributing a preset weight to each set, then transmitting to-be-classified data into the random forest which is constructed in the step 1) for performing classification, and adjusting the weight according to a classification result and a predication result, if the classification predication result of the set does not accord with the actual result, increasing the weight of the set, and if the classification predication result accords with the actual result, reducing the weight of the set; and a third step, performing classification on the to-be-classified data according to the adjusted weight of each set until the classification result accords with the predication result.

The invention discloses a method and module for extracting and interpreting information of a remote-sensing image, and provides a geographical element object based automatic interpreting method based on SRMMHR and random forest. The module extracts and interprets the information of the remote-sensing image by the adoption of the method. The method and module for extracting and interpreting the information of the remote-sensing image can achieve effective segmentation of the remote-sensing image, automatic optimization of characteristics, and automatic establishment of a classification rule set; a means which is little in manual intervention, high in degree of automation and high in interpreting precision is provided for extracting and interpreting the information of the remote-sensing image.

The invention discloses improved random forest algorithm based system and method for software fault prediction. The system comprises a data processing layer, a prediction model building layer and a fault predication layer. The method includes calculating a software project attribute set used for acquiring a training model to acquire a training data set of a software prediction model, and performing equalization to the training data set; building a prediction model according to an improved random forest algorithm; screening the model according to performance limiting of accuracy rate and recall ratio; and predicting a software project according to attribute set information of the to-be-predicted software project and a trained prediction model and displaying prediction results and the prediction model. The improved random forest algorithm based system and method for software fault prediction have the advantages of high prediction accuracy rate, performance stability and high execution efficiency, can evaluate whether a final software product reaches specified quality or meets expectation of a user or not, and can guide developers to formulate distribution strategies of software testing and formal verification resources.

The invention discloses an abnormal behavior discovery method and system based on big data machine learning. The abnormal behavior discovery method disclosed by the invention comprises the following steps: carrying out pretreatment on the original security log data; extracting characteristic data from a pretreatment result; clustering the characteristic data, and determining an abnormal behavior library and a normal behavior library; acquiring new behavior sample data in the security log data, comparing with the normal behavior library and the abnormal behavior library, determining a new behavior to be a normal behavior or an abnormal behavior, and updating the normal behavior library or the abnormal behavior library with the new behavior sample data; and repeating the previous step, when the normal behavior library and the abnormal behavior library have enough normal behavior and abnormal behavior sample data, training a random forest model with sample data in the normal behavior library and the abnormal behavior library, and judging the abnormal behavior by utilizing the random forest model obtained through training. By adopting the scheme of the invention, the problem that quantity of label-containing samples in an initial stage is too low is solved, judging accuracy rate is improved, and misjudgement condition is effectively prevented from occurring.

In one embodiment of the present invention, a graphics processing unit (GPU) is configured to detect an object in an image using a random forest classifier that includes multiple, identically structured decision trees. Notably, the application of each of the decision trees is independent of the application of the other decision trees. In operation, the GPU partitions the image into subsets of pixels, and associates an execution thread with each of the pixels in the subset of pixels. The GPU then causes each of the execution threads to apply the random forest classifier to the associated pixel, thereby determining a likelihood that the pixel corresponds to the object. Advantageously, such a distributed approach to object detection more fully leverages the parallel architecture of the PPU than conventional approaches. In particular, the PPU performs object detection more efficiently using the random forest classifier than using a cascaded classifier.

The invention discloses a short period load prediction method based on kernel principle component analysis and a random forest. The a short period load prediction method comprises the following steps of: (1) analyzing and selecting data influencing load prediction precision of a day to be predicted in an operational electric power system, and preliminarily constructing training and prediction sample sets; (2) utilizing kernel principle component analysis to carry out dimensionality reduction on training sample data; (3) utilizing a random forest model to train the training sample data after the dimensionality reduction, and obtaining the random forest model after the training; and (4) inputting prediction sample data into the random forest model after the training, and carrying out short period load prediction of the day to be predicted. The short period load prediction method has the advantages that the kernel principle component analysis and the random forest model are combined for carrying out short period load prediction on the electric power system, the prediction precision, efficiency and data rationality are improved.

The invention discloses a strip steel surface area type defect identification and classification method which comprises the following steps: extracting strip steel surface pictures in a training sample database, removing useless backgrounds and keeping the category of the pictures to a corresponding label matrix; carrying out bilinear interpolation algorithm zooming on the pictures; carrying out color space normalization on images of the zoomed pictures by adopting a Gamma correction method; carrying out direction gradient histogram feature extraction on the corrected pictures; carrying out textural feature extraction on the corrected pictures by using a gray-level co-occurrence matrix; combining direction gradient histogram features and textural features to form a feature set, which comprises two main kinds of features, as a training database; training the feature data with an improved random forest classification algorithm; carrying out bilinear interpolation algorithm zooming, Gamma correction, direction gradient histogram feature extraction and textural feature extraction on the strip steel defect pictures to be identified in sequence; and then, inputting the feature data into an improved random forest classifier to finish identification.

The present invention discloses a computer vision-based express parcel violent sorting identification method. The method comprises the following steps of: a depth camera-based pose estimation: a human pose estimation problem is converted into a problem of classifying depth image pixels captured by a depth camera, and human body pose estimation is realized by using a random forest method; human body three-dimensional pose relative spatial-temporal feature extraction: relative spatial-temporal positions of geometric elements of points, lines and surfaces formed by joints in three-dimensional poses and the measures of the change of the relative spatial-temporal positions are extracted and are adopted as the feature representations of the poses; and recurrent neural network-based violent sorting identification: modeling training is performed on the pose spatial-temporal relative features which are continuous in time and are extracted from the human body three-dimensional poses through a long and short memory model (LSTM), so that the identification of express parcel violent sorting behaviors can be realized.

The present application discloses a method for updating a classification model, comprising: obtaining incremental data within a predetermined period of time from a server providing the user behavior data as a training sample set; determining the number of newly added decision trees; According to the training sample set, the random forest algorithm is used to generate the newly added number of decision trees; the decision trees included in the classification model and the newly generated decision trees are sorted according to the classification effect, and the sequence position is selected from them. High-level, the predetermined number of decision trees; summarizing the selected decision trees to obtain an updated classification model. The present application also provides a device for updating a classification model. With the method provided in this application, since it is not necessary to perform training based on the full amount of data, but to use the incremental update method on the basis of the original classification model, it can improve the efficiency of model training and achieve rapid response to business.

The invention relates to a method for identifying multi-class facial expressions at a high precision based on Haar-like features, which belongs to the technical field of computer science and graphic image process. Firstly, the high-accuracy face detection is achieved by using the Haar-like features and a series-wound face detection classifier; further, the feature selection is carried out on the high-dimension Haar-like feature by using the Ada Boost. MH algorithm; and finally, the expression classifier training is carried out by using the Random Forest algorithm to complete the expression identification. Compared with the prior art, the method can reduce training and identifying time while increasing the multi-class expression identification rate, and can implement the parallelization conveniently to increase the identification rate and meet the requirement of real-time processing and mobile computing. The method can identify the static image and the dynamic image at a high precision, is not only applicable to the desktop computer but also to the mobile computing platforms, such as cellphone, tablet personal computer and the like.

The invention provides a mobile data traffic package recommendation algorithm based on user historical data according to data mining analysis technology. The mobile data traffic package recommendation algorithm comprises the following steps of: 1) a target user finding period comprising the processes of a, acquiring a processed generated data set which comprises a training set and a prediction set, b, executing a random forest classification algorithm for finding a latent data traffic package improving user as a target user, and c, ending; 2), a data traffic package recommendation period comprising the process of a, acquiring a processed generated prediction set, b, executing a K-means clustering algorithm for obtaining a slightly similar user cluster, c, obtaining the target user obtained in the process 1)-b, d, executing a TopN recommendation algorithm on the target user in a same cluster according to a similarity function of the user, and e, ending. The mobile data traffic package recommendation algorithm is used for finding the latent user with a latent data traffic improvement requirement according to data mining technology and executing a recommended plan on the user. Compared with a traditional method, the mobile data traffic package recommendation algorithm has advantages of higher accuracy, higher efficiency, simple realization, low cost, etc.

The invention discloses a travel time prediction method for optimizing an LSTM neural network through a particle swarm optimization algorithm, and the method comprises the following steps: S1, collecting travel time data, performing data normalization, and dividing the data into a training set and a test set proportionally; S2, optimizing each parameter of an LSTM neural network prediction model by using the particle swarm optimization algorithm; S3, inputting the parameters, optimized through the particle swarm optimization algorithm, and the training set, and performing the iterative optimization of the LSTM neural network prediction model; S4, predicting the test set through the trained LSTM neural network model, and evaluating a model error. The method is quick in optimization. Compared with a random forest, SVM and KNN in the traditional prediction algorithm, the method of the invention has the least mean square error and square error for the data prediction, and the model reducesthe calculation burden, so the method shows better prediction performance.

The invention provides a method for detecting gestures. The method includes steps of detecting a predefined motion mode to determine a region-of-interest to be detected; detecting in the region-of-interest according to a scheme of a multi-scale sliding window, extracting local mean features from skin color membership images based on window images at first, classifying the features through a pre-trained Ada-Boost classifier, further extracting feature points on the basis of a gray-scale image of the window images processed via the Ada-Boost classifier, and then classifying the point-pair features through a random forest classifier; and clustering different types of target gesture windows and outputting accurate positions and shapes of the gestures. The invention further provides a system for detecting the gestures. The method and the system for detecting the gestures are simple, rapid and stable in implementing, and have the advantages of real-time performance, interference resistance, high identification precision and the like.

The invention discloses a TV shopping commodity recommendation method based on a classification algorithm. The TV shopping commodity recommendation method comprises the steps of: converting a prediction problem into a classification problem by utilizing logistic regression and a random forest, namely, predicting that purchasing behaviors of a user about a commodity can be divided into two categories: purchasing and not purchasing; extracting features from commodity information, user information and a user behavior record as input, and taking a prediction score of the user as output, thereby forming a function; and training a model by adopting a linear regression method, and converting the problem into a training classifier problem. The TV shopping commodity recommendation method does not carry out prediction and calculation on the basis of a heuristic rule, but carries out prediction on the basis of data analysis and statistics as well as machine learning for training model; and new users and new commodities can be calculated and predicted quickly as long as the model is trained.

The invention provides a criminal identification and forecast method. The method adopts a data pre-processing method in data mining; aiming at criminal information such as data, street address, criminal police zone, week, criminal type, criminal description and sentence processing, attribute reconstruction, feature extraction and feature selection are performed, the correlation between the criminal information is mined, a characteristic factor with maximum difference is generated, and the correlation between the characteristics factor and a criminal result, namely the criminal type is generated; and then a model integrating Gaussian Naive Bayes, a neural network, Logistic regression, regularized regression, K neighbor, random forest, a support vector machine and an XGBoost learning algorithm is built to obtain an element classifier based on a weighted voting classifier having highlight classification and favorable clustering effect, reconstructed data is analyzed, processed and identified, a criminal condition of a city in future is forecasted, an individual criminal map of the city is drawn, and the effects of promoting and regulating city public security and management are further achieved.

The invention discloses an identification method of a harassment number. The identification method of the harassment number comprises the steps of selecting a plurality of harassment numbers and non-harassment numbers which are confirmed; calculating communication behavior indexes of the harassment numbers and non-harassment numbers within a period of time; then forming a training sample set by using the harassment numbers and non-harassment numbers as well as the communication behavior indexes thereof so as to build a random forest classification model, wherein the input of the random forest classification model is the communication behavior index of each user number and the output thereof is a prediction probability of judging each user number as a harassment number or non-harassment number by all the decision-making trees; and inputting the communication behavior index of a to-be-identified number within a period of time into the random forest classification model, calculating the prediction probability of judging the to-be-identified number as a harassment number or non-harassment number by all the decision-making trees, and accordingly determining whether the to-be-identified number is a harassment number. The identification method of the harassment number is belongs to the technical field of network communication, can efficiently identify harass numbers from massive traffic data of the present network by making full use of calling features of calling and called numbers.

The invention discloses a fall-down behavior real-time detection method based on a depth image. The method comprises steps of depth image obtainment, human body image identification, depth difference feature extraction, human body part analysis, articulation point extraction, height feature extraction and fall-down behavior detection. Based on a depth image, a specific depth difference feature is selected on an identified human body image; a random forest classifier is used for analyzing human body parts; a human body is divided into a head part and a trunk part; articulation points are detected, and then a height feature vector is extracted; and a support vector machine classifier is used for detecting whether a detected object is in fall-down state. The invention provides the fall-down behavior detection method, improves the operation speed, and achieves timeliness of fall-down behavior detection. The depth image is utilized for fall-down behavior detection. On one hand, the method is free of influences of illumination and can be operated in an all-weather manner, and on the other hand, personal privacy can be protected compared with a colorful image. Only one depth sensor is required in terms of hardware support, and the advantage of low cost is achieved.

A method, apparatus and computer program product are provided to not only predict an impending fault, but also to avoid the occurrence of the fault such that continued operations are permitted with a reduced likelihood of the occurrence of the fault. In this regard, a plurality of features are provided to at least one model, such as a random forest classification model. The plurality of features include features representative of at least one prior operational sequence as well as features representative of at least one upcoming operational sequence. The plurality of features are then processed with at least one model to determine a likelihood of a fault during the upcoming operational sequence. The method also alters the characteristics of the upcoming operational sequence without requiring maintenance of the equipment to thereby permit the upcoming operational sequence to be conducted with a reduced likelihood of the fault.

The invention belongs to the technical field of image processing, discloses a method for identifying and positioning power transmission line insulators in unmanned aerial vehicle aerial images, and solves the problems in the prior art that the detection precision of an identification algorithm of the insulators is not high, the robustness is low, and the identification algorithm is easy to be affected by sample number. A group of Gabor wavelet basis with different sizes and different directions and training sample images are taken as convolutions so as to form a group of characteristic vectors which accurately describe sample image texture characteristics. A random forest machine learning algorithm with a semi-supervised learning mode is used to train sample data sets of the known category and the unknown category so as to obtain an insulator identification model. Through the mode from left to right and from top to bottom, a detection window with the same size as the training sample traverses the input images with different sizes. The detection window combining the identification model detects and positions the positions of the insulators in the input images with different sizes. And finally the accurate positions of the insulators in the input image with the original size are determined by using a non-maximum inhibition method.