1499 results about "Logistic regression" patented technology

A method of building predictive statistical models provides a dedicated aggregation module for each transactional record source. Each aggregation module aggregates the transactional records using a neural network function to produce a scalar output which can then be input to a traditional modeling function, which may employ either logistic regression, neural network, or radial basis function techniques. The output of the aggregation modules can be saved, and updated aggregation values can be updated by processing new transaction records and combining the new transaction values with the previous output values using a blending function. Parameters of the neural network in the aggregation module may be calculated simultaneously with the parameters of the traditional modeling module.

The invention provides methods, systems, and computer program products for short term probability forecasting of selected weather-related events. These embodiments are adaptable for any geographical region that can be identified and for which a reasonable number of data points exist. The inventive method uses a data set of n observations of m parameters, where the parameters may be statistically correlated. A Principal Component Analysis may be performed, with the data set as input, to provide a reduced set of principal components that are uncorrelated and account for most of the variance in the input data set. An orthogonal transformation may be performed on the reduced set of principal components to provide a rotated set of principal components that are aligned with the corresponding parameters in the input data set. Finally a logistic regression may be performed on the rotated set of principal components to derive an S-shaped predictive equation for the probability of a binary weather-related event of interest. An illustrative embodiment of the invention is given for forecasting the probability for the number of lightning flashes exceeding a selected value, for the western United States climatological area.

Determining whether two merchant location database entries are describing the same merchant location. A subject merchant location database entry and comparison candidate merchant location database entries include a DBA name field, a street address field, and one or more additional descriptive fields descriptive of one or more predetermined characteristics of the respective merchant location. The subject merchant location database entry is compared to a set populated with candidate merchant location database entries, candidates having a predetermined minimum textural similarity with the subject merchant location database entry on the basis of each entry's DBA name field or street address field. The subject merchant location database entry is compared with each of the candidate database entries on the basis of the one or more additional descriptive fields, and a logistic regression is performed using the results of the comparing, in order to calculate a probability that the database entries refer to the same merchant location.

The invention belongs to the field of depth learning, and provides a rail surface defect detection method based on the depth learning network, aiming at solving various problems existing in the priorrail detection methods. The depth learning method first automatically resets the input rail image to 416*416, and then extracts and processes the image. Image extraction mainly by Darknet-53 model complete. The processing output is mainly accomplished by the FPN-like network model. Firstly, the rail image is divided into cells. According to the position of the defects in the cells, the width, height and coordinates of the center point of the defects are calculated by dimension clustering method, and the coordinates are normalized. At the same time, we use logistic regression to predict the fraction of boundary box object, use binary cross-entropy loss to predict the category contained in the boundary box, calculate the confidence level, and then process the convolution in the output, up-sampling, network feature fusion to get the prediction results. The invention can accurately identify defects and effectively improve the detection and identification rate of rail surface defects.

The present invention is in the field of Healthcare Claims Fraud Detection. Fraud is perpetrated across multiple healthcare payers. There are few labeled or “tagged” historical fraud examples needed to build “supervised”, traditional fraud models using multiple regression, logistic regression or neural networks. Current technology is to build “Unsupervised Fraud Outlier Detection Models”.Current techniques rely on parametric statistics that are based on assumptions such as outlier free and “normally distributed” data. Even some non-parametric statistics are adversely influenced by non-normality and the presence of outliers.Current technology cannot represent the combined variable values into one meaningful value that reflects the overall risk that this observation is an outlier. The single value, the “score”, must be capable of being measured on the same scale across different segments, such as geographies and specialty groups. Lastly, the score must substantially, monotonically rank the fraud risk and give reasons to substantiate the score.

New and efficient protocols are provided for privacy-preserving machine learning training (e.g., for linear regression, logistic regression and neural network using the stochastic gradient descent method). A protocols can use the two-server model, where data owners distribute their private data among two non-colluding servers, which train various models on the joint data using secure two-party computation (2PC). New techniques support secure arithmetic operations on shared decimal numbers, and propose MPC-friendly alternatives to non-linear functions, such as sigmoid and softmax.

The invention provides a human body behavior identification method based on an inertial sensor. The method comprises the steps of acquiring human body behavior data of testees by means of the inertial sensor, conducting sliding window segmentation on the acquired human body behavior data, conducting feature extraction on a triaxial accelerated speed subsequence and a triaxial angular speed subsequence which are obtained after sliding window segmentation, conducting feature fusion on a feature vector to form a sample set of the human body behaviors of the testees, conducting feature selection on the sample set by means of the least correlated maximum redundant algorithm and the Bayes regularization sparse polynomial logistic regression algorithm, obtaining the classification feature vectors of all human body behaviors of all the testees, obtaining a classifier model of each human body behavior by means of a fuzzy least square support vector machine, and obtaining a human body behavior identification result after the human body behavior data are tested by means of the fuzzy least square support vector machine. By means of the human body behavior identification method, self-adaptation and identification efficiency can be improved.

Embodiments of the present invention provide a method implemented by a computer program for detecting and identifying valve failure in a reciprocating compressor and further for predicting valve failure in the compressor. Embodiments of the present invention detect and predict the valve failure using wavelet analysis, logistic regression, and neural networks. A pressure signal from the valve of the reciprocating compressor presents a non-stationary waveform from which features can be extracted using wavelet packet decomposition. The extracted features, along with temperature data for the valve, are used to train a logistic regression model to classify defective and normal operation of the valve. The wavelet features extracted from the pressure signal are also used to train a neural network model to predict to predict the future trend of the pressure signal of the system, which is used as an indicator for performance assessment and for root cause detection of the compressor valve failures.

The invention discloses a goods recommendation method based on scores and user behaviors. First of all, a latent factor model is established for user score data, goods are automatically clustered, latent classes or feature factors are found, user interest is decomposed into preference degrees of the multiple latent classes, the goods are expressed by use of weights comprising latent features, and the scores of the users for the goods are inner products of the user interest and the goods. Then for the purpose of solving the score data sparsity problem, by use of the user behaviors, negative samples are introduced, the features are extracted, and a possibility that the users buy the goods is estimated through a logic regression model. Finally, candidate sets of the two are combined and weighed for ordering, and top goods are recommended to the users. According to the invention, diversified interest of the users is discovered from the single scores by use of the latent factor model, information of the multiple features of the goods is mined, the method better accords with actual application, the negative samples are introduced, distinctiveness of the user interest is enabled to be larger, the quality of a recommendation result is higher, demands of the users can be better satisfied, and the method can be applied to recommending the goods.

Illness signatures are mathematically characterized by entrainment relationships among multiple time series representations of physiological processes. Such characteristics include time and phase lags, window lengths for optimum detection, which time series are most entrained with each other, the degree of entrainment relative to the rest of the large database, and the concordance or discordance of the time-varying changes. These optimum disease-specific characteristics can be determined, for example, from large, clinically well-annotated databases of time series representations of physiological processes during health and illness. These characteristics of the entrainment relationships among multiple time series representations of physiological processes are used to make mathematical and statistical predictive models using multivariable techniques such as, but not limited to, logistic regression, nearest-neighbor techniques, neural and Bayesian networks, principal and other component analysis, and others. These models are quantitative expressions that transform measured characteristics to the probability of an illness, or p(illness).

Embodiments of the invention provide a text relevance calculating method and a device thereof. The method comprises the following steps: receiving a first character string and a second character string; calculating a text relevance characteristic value of the first character string and the second character string, and calculating a semantic relevance characteristic value of the first character string and the second character string; fitting the text relevance characteristic value and the semantic relevance characteristic value into a relevance characteristic value of the first character string and the second character string based on the logistic regression model. The text relevance calculating method and the device thereof increase the precision of relevance judgment, save storage space and reduce the cost.

The invention relates to a face image gender recognition method based on stack type sparse self-coding, and belongs to the field of image recognition, machine learning, and computer vision. A training process of the method includes image graying, histogram equalization, geometric correction, image normalization, the training of a sparse self-coding model, logic regression classifier training, a fine tuning model, and model fusion of face standard databases FERET and CAS-PEAL-R1, and a prediction process comprises the capturing of natural scene images by a camera, image graying, histogram equalization, face detection, geometric correction, image normalization, the prediction by employing a stack type sparse self-coding model, and result marking. According to the method, the problem of face gender recognition is solved by employing the stack type sparse self-coding model, combination characteristics of the images can be learned layer by layer, original signals can be better represented in an abstract manner, characteristics extracted by a hiding unit are further adjusted by the adoption of fine tuning, and the recognition accuracy is higher.

The present invention provides a solution for predicting an advertisement click-through rate. The solution comprises: acquiring characteristic related information of multiple characteristic types related to multiple past delivered advertisements in a predetermined past time period; performing cross combination on characteristic related information of at least two characteristic types of each past delivered advertisement, to determine multiple cross characteristic sets, and calculating to determine cross characteristic identifiers separately corresponding to the multiple cross characteristic sets; extracting an advertisement display quantity and an advertisement click quantity corresponding to each cross characteristic set, and calculating to determine an advertisement click-through rate corresponding to each cross characteristic set, so as to use the advertisement click-through rate as a cross characteristic value; performing training on a logistic regression model based on the cross characteristic identifiers and the cross characteristic values separately corresponding to the multiple cross characteristic sets, and calculating to determine a model training parameter; and performing prediction calculation on an advertisement click-through rate of a to-be-predicted advertisement based on the model training parameter. According to the solution, more reliable training data is provided for prediction calculation of an advertisement, so that accuracy of a prediction calculation result of an advertisement click-through rate is ensured.

A robust classification method for cancer detection from mass spectrometry data includes inputting the mass spectrometry data, preprocessing the spectrometry data, conducting robust feature selection, generating predictions for the test data sets using multiple data classifiers, the multiple data classifiers including artificial neural networks, support vector machines, weighted voting on data patterns, classification and regression trees, k-nearest neighbor classification, and logistic regression, and constructing and validating a meta-classifier by combining individual predictions of the multiple data classifiers to generate a robust prediction of a phenotype. The test data sets are used exclusively for validation of the meta-classifier.

A method for analysing incoming data, comprising the steps of processing the incoming data in segments to output a sequence of segment types by extracting one or more properties of an incoming data segment and forming an Unknown Property Vector for each segment of data in the incoming data, and processing the sequence of segment types to identify events in the incoming data. The sequence of segment types is determined, for each segment, by reference to a set of Reference Property Vectors that are relevant to the Unknown Property Vector. This may involve application of first and / or second and / or further functions to identify at least a first subset of Reference Property Vectors that are relevant to the Unknown Property Vector. Alternatively, a logistic regression algorithm, derived using clustering or classification methods for identifying candidate vectors, may be used.

Embodiments of the invention present an approach to conversion rate estimation which relies on using past performance observations along user, publisher, and advertiser data hierarchies. More specifically, embodiments of the invention model the conversion event at different select hierarchical levels with separate binomial distributions and estimate the distribution parameters individually. It is shown how to combine these individual estimators using logistic regression to identify conversion events accurately. Embodiments of the invention can also handle many practical issues, such as data imbalance, missing data, and output probability calibration, which render this estimation problem more difficult for a real-world implementation of the approach.

The invention discloses an improved high-resolution remote sensing image classification method based on deep learning. On the basis of the deep learning theory, a seven-layer convolutional neural network is designed; a high-resolution remote sensing image sample is inputted into the network to carry out network training and last two full connection layers obtained by learning are outputted as twodifferent high-level features of the remote sensing image; dimension reduction is carried out by using a principal component analysis for the output of the fifth pooling layer of the network, whereinthe result after dimension reduction is used as a third high-level feature of the remote sensing image; the three kinds of high-level features are fused in series; and then an effective logistic-regression-based classifier is designed to classify the remote sensing image. According to the invention, feature extraction is carried out on the high-resolution remote sensing image based on the deep learning theory and the features obtained by learning have high expressive force and robustness. Besides, the extracted high-level features are fused and the fused feature is inputted into the logistic regression classifier, so that the good classification result is obtained.

A readmission risk prediction model is generated and used for identifying patients having elevated risk of readmission and determining inpatient treatment and outpatient activities based on readmission risk. Readmission risk prediction models may be generated for a variety of different clinical conditions using logistic regression techniques. When a patient is admitted to a hospital, the patient's condition is identified and a corresponding readmission risk prediction model is employed to identify the patient's risk of readmission. The readmission risk may be presented to a clinician and employed to recommend interventions intended to treat the patient and reduce the probability of readmission for the patient. The patient's readmission risk may also be calculated after the patient has been discharged and used for planning outpatient activities for the patient.

The invention relates to a state recognition and prediction method for a spindle characteristic test bench based on deep learning, which comprises the steps of collecting vibration signals in the operating process of the spindle characteristic test bench, performing normalization processing on the vibration signals, performing noise reduction processing on the normalized vibration signals by adopting EEMD (Ensemble Empirical Mode Decomposition) to obtain IMF components, and reconstructing the obtained IMF components to form restored signals; enabling the restored signals to serve as input samples of a CNN, performing feature extraction on the restored signals to obtain feature vectors, carrying out CNN feature learning on the feature vectors to obtain training feature samples; coding timeinformation for the training feature samples through a multi-layer LSTM (Long Short Term Memory), carrying out classification through Softmax logistic regression to obtain prediction feature samples,and realizing prediction for the operating state; and performing Softmax logistic regression through the training feature samples and the prediction feature samples, carrying out classification on a logistic regression layer so as to judge the fault type of a rotor rotation test bench system, and realizing state recognition. The state recognition and prediction method has fast response performanceand tracking performance.

The invention discloses a time sequence classification early warning method for a storage device. The method comprises the steps of collecting storage device parameters in real time; cleaning data; performing ARIMA time sequence analysis; and performing logistic regression analysis and early warning mechanism output. Under the background of a big data environment, time sequence prediction analysisis performed by adopting an ARIMA model according to historical data and hard disk SMART information obtained by statistics; the correlation between a SMART eigenvalue and a fault rate of the storagedevice is analyzed; and an eigenvalue more suitable for a Logistic model is selected out to perform classification prediction. A machine learning method is adopted for predicting the fault rate of the storage device, so that the problems of classification singleness and low early warning intensity in final prediction of the storage device are solved, the defects of hysteresis, low accuracy, pooractual early warning effect and difficult application to the big data environment for a disk early warning mechanism in the prior art are overcome, the occurrence probability of each early warning intensity can be predicted, and an effective solution is provided for real-time operation maintenance and monitoring in a data center environment.

The invention discloses a detection method and a detection system for fraud access to a business to business (B2B) platform based on data mining. The detection method includes dividing the information of clients into static information and dynamic information, detecting the static information by means of a data mining method of an association analysis, detecting the dynamic information by means of a data mining method of a logistic regression classification model, comprehensively calculating warning values obtained from two data mining methods, grading the clients whose warning values exceed the threshold value, judging the access clients who are graded into a specific grade to be fraud visitors, and listing the fraud visitors into a blacklist information base of fraud clients. The detection system comprises a client information processor, a fraud analysis processor and a front-end display processor. According to the detection method and the detection system for the fraud access to the B2B platform based on the data mining, by means of characteristics of the B2B e-commerce platform, on the basis of multi-dimensional data of client information data, client accessing behaviors and the like, the detection method and the detection system detect behaviors of the fraud access to the B2B e-commerce platform by introducing the data mining technology to model, and the problem that the fraud accesses are difficult to detect caused by the fact that transaction behaviors can't be monitored in the industry is solved.

The invention belongs to the field of image processing, and particularly relates to an automatic extraction method of regions of interest in three-dimensional breast full-volume images (ABVS). The method comprises the following steps: processing the continuous cross section two-dimensional images in three-dimensional ABVS images by using a maximum direction-based phase information method to obtain the candidate regions of interest on each cross section image; removing the unrelated regions according to the prior knowledge such as the continuity and position characteristic of breast tumor on the two-dimensional cross section images; obtaining the shape and texture features of the residual suspected tumor regions, inputting the shape and texture shapes to a two-valued logistic regression classifier to obtain the probability of each region becoming tumor and selecting the region with the maximum probability as the tumor region; obtaining the minimum ellipsoid comprising the region of interest according to the selected region to serve as the region of interest. The automatic extraction method provided by the invention can be used for realizing the automatic extraction of tumor regions of interest in the three-dimensional ABVS images, obtaining the correct positions of tumor, decreasing the workload of the manual operation and providing important reference to further tumor detection.

The invention relates to photo cheat-preventing living body detecting method and system which can be used for identity authentication application based on human face recognition. The system comprises an intermediate-frequency filter, a frequency-domain converter and a classifier, wherein the intermediate-frequency filter can adopt a DoG (Difference of Gaussian) filter and is used for carrying out filtering pretreatment on an image to obtain intermediate frequency band information; the frequency-domain converter can adopt a Fourier converter and is used for extracting Fourier transform characteristics from the pretreated two-dimensional image; and the classifier can adopt a logistic regression classifier and is used for judging whether the image acquired from the identity authentication is a real human face or a photo human face. Proved by test results, the method and the system can favorably solve the problem of photo cheat in the identity authentication based on human face recognition under the conditions of no addition of additional auxiliary equipment, no need of active matching of a user, simple realization, less calculated amount and independent function.

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 advertisement click-through rate prediction method based on multi-dimensional feature combination logical regression. The method comprises the first step that feature information of a hierarchical structure of the user hierarchy, feature information of a hierarchical structure of the media hierarchy and feature information of a hierarchical structure of the advertisement hierarchy are extracted from the obtained click-through rate data respectively; the second step that multi-dimensional combination is carried out on the feature information of the hierarchical structure of the user hierarchy, the feature information of the hierarchical structure of the media hierarchy and the feature information of the hierarchical structure of the advertisement hierarchy, three-to-three combination is carried out on one-dimensional feature information in the feature information to obtain a three-dimensional feature combination, and a feature vector combined by the three-dimensional feature information is formed to represent a user cluster; the third step that the second step is carried out repeatedly and a learning set of the feature vector combined by the three-dimensional feature information is obtained; the fourth step that the learning set obtained in the third step is used for training and testing a logical regression model, and the logical regression model is used for predicting the advertisement click-through rate.

The invention discloses a filtering method of recursive Chinese junk E-mail, which is based on Logistic. The method comprises the following steps: first, analyzing E-mails, extracting E-mail titles, E-mail main bodies and accessory relative information, second, segmenting words for version information which is extracted, third, accounting word frequencies of entries in E-mails, calculating weights of words through utilizing TF-IDF pattern, presenting the E-mail to be characteristic vector which is weighted, fourth, utilizing an LIBLINEAR tool kit to exercise the sample of the E-mail to get an Logistic recursive module, fifth, utilizing the Logistic recursive module to classify for new E-mails, getting the probability value whether the E-mails which are got are junk E-mails. The utility which utilizes the Logistic recursive module has the advantages of simple module, little amount of parameter, and high classifying accuracy in a data set whose text number and characteristic number are both bigger, the accuracy and efficiency of filtering junk E-mails are improved through dimension reduction and improved characteristic value calculating method, and meanwhile, the problem of choosing module exercise parameter which is faced in filtering junk E-mails is effectively solved.

Systems, methods, and apparatuses are disclosed for identifying road signs along a roadway segment from vehicle probe data. Probe data is received from vehicle sensors at a road segment, wherein the probe data includes observed static objects along the road segment. Road signs are identified within the observed static objects of the probe data using a logistic regression algorithm. The geographic location of the identified road signs are determined using a linear regression algorithm.

The present invention provides systems and methods for signal detection and enhancement. The systems and methods utilize one or more discriminative classifiers (e.g., a logistic regression model and a convolutional neural network) to estimate a posterior probability that indicates whether a desired signal is present in a received signal. The discriminative estimators generate the estimated probability based on one or more signal-to-noise ratio (SNRs) (e.g., a normalized logarithmic posterior SNR (nlpSNR) and a mel-transformed nlpSNR (mel-nlpSNR)) and an estimated noise model. Depending on the resolution desired, the estimated SNR can be generated at a frame level or at an atom level, wherein the atom level estimates are utilized to generate the frame level estimate. The novel systems and methods can be utilized to facilitate speech detection, speech recognition, speech coding, noise adaptation, speech enhancement, microphone arrays and echo-cancellation.