62 results about "Factor selection" patented technology

The invention comprises transformation of a section of a data block independently of the transformation of separate or overlapping data blocks to determine a property related to the original matrix, where each of the separate or overlapping data blocks are derived from an original data matrix. The transformation enhances parameters of a first data block over a given region of an axis of the data matrix, such as signal-to-noise, without affecting analysis of a second data block derived from the data matrix. This allows for enhancement of analysis of an analyte property, such as concentration, represented within the original data matrix. A separate decomposition and factor selection for each selected data matrix is performed with subsequent score matrix concatenization. The combined score matrix is used to generate a model that is subsequently used to estimate a property, such as concentration represented in the original data matrix.

The invention provides for transformation of a section of a data block independently of the transformation of separate or overlapping data blocks to determine a property related to the original matrix, where each of the separate or overlapping data blocks are derived from an original data matrix. The transformation enhances parameters of a first data block over a given region of an axis of the data matrix, such as signal-to-noise, without affecting analysis of a second data block derived from the data matrix. This allows for enhancement of analysis of an analyte property, such as concentration, represented within the original data matrix. In a first embodiment of the invention, a separate decomposition and factor selection for each selected data matrix is performed with subsequent score matrix concatenization. The combined score matrix is used to generate a model that is subsequently used to estimate a property, such as concentration represented in the original data matrix. In a second embodiment, each data matrix is independently preprocessed. Demonstration of the invention is performed through glucose concentration estimation from noninvasive spectra of the body.

The invention discloses a risk assessment algorithm for an information system. According to the GB / T20984-2007 standard, a correlation between the assessment factors of assets, the assessment factors of vulnerability and the assessment factors of threats of the information system is established, a safety assessment indicator system is achieved, and 24 pairs of risk relations are achieved. The 24 pairs of achieved risk relations are substituted into a formula (1), an asset comprehensive value A is obtained through calculation. According to asset comprehensive value A and a vulnerability value V, the comprehensive value F of the loss caused by security events is worked out. According to the vulnerability value V and a threat value T, a security event possibility comprehensive value L is worked out. The comprehensive value F of the loss caused by the security events and the security event possibility comprehensive value L are substituted into a formula (2), and then a risk comprehensive value R is worked out and obtained. The risk assessment algorithm for the information system can eliminate the influence caused by the facts that assessment factor selection is unreasonable and risk correlation analysis cannot objectively reflect the system state, and improve the objectivity and the accuracy of risk assessment.

The present invention discloses a method for selecting main substation capacity and an optimal station address of a transformer substation and belongs to the technical field of distribution network planning in the power industry. According to the present invention, a method for selecting the station address of the transformer substation comprises: selecting the optimal station address of the transformer substation according to factors such as an area load condition, load distribution, a geographic position and the like; and a method for determining the substation capacity comprises: selecting and determining the main substation capacity of the transformer substation according to the load distribution and the future development condition by considering conditions such as investment cost and the like. According to the present invention, based on the actual situations, a load is predicted and according to the load prediction condition and the actual situations of environmental geographical factors and the like, a transformer substation layout principle is combined, factors such as initial investment cost, operation cost, maintenance cost and the like of a transformer substation planning scheme are comprehensively considered and related constraint conditions such as a capacity-load ratio and the like are considered so as to determine the optimal substation capacity and station address; and the method considers more comprehensive actual factors of a distribution network transformer substation locating and sizing plan, so that the planning scheme is closer to the actual situations, the method is more reasonable and practicality is higher.

The invention provides a traffic flow monitoring and predicting system which comprises a monitoring device and a predicting device connected with the monitoring device, wherein the predicting device comprises a collection module, a data preprocessing module, a data classifying module, a stability check module, a correlation coefficient calculation module, a threshold setting module, a space-time correlation coefficient matrix generation module, a historical correlation coefficient matrix generation module, a prediction factor selection module and a prediction model construction module which are sequentially connected. The traffic flow monitoring and predicting system is relatively high in prediction accuracy and the constructed prediction model has pertinence.

The invention provides a credit overdue prediction method and system fused with machine learning, and the method comprises the steps: collecting a plurality of credit factor data, carrying out the preprocessing, carrying out the calculation and sorting of the importance of the credit factor data in a preprocessing result, and deleting redundancy, and obtaining the selected credit factor data; andconstructing a training sample based on the credit factor data, establishing and training a credit overdue prediction model by using LSTM based on the training sample, determining an optimal parameter, and performing credit overdue prediction after the optimal model is obtained. According to the invention, credit factor data is widely collected to improve comprehensiveness of credit overdue prediction; the missing training data is classified to improve the data quality; the class imbalance condition of the user is processed by using an oversampling method, and data distribution is balanced; all factors influencing credit expiration is sorted, and redundancy is eliminated, and then the reasonability of factor selection is improved; and a credit overdue prediction model is comprehensively established based on bidirectional LSTM in combination with timing sequence factors, optimal model parameters are determined through S-fold intersection, and the optimal model quality is improved.

The invention relates to a statistical downscaling model forecasting factor screening method. The method comprises the steps that 1 a positional relationship graph between a weather station and a gridin a river basin is established; 2 the correlation among measured rainfall of each station, temperature and other predictand and a reanalysis data atmospheric circulation factor is deduced, includingthe correlation with the grid of the weather station and the correlation with surrounding grids, and appropriate factors are selected according to a set correlation coefficient threshold; 3 the change of a criterion certainty coefficient in a calibration period on monthly, seasonal and annual scales is analyzed; and 4 according to different scales and different factor selection methods in the step 3, a model is analyzed and tested.

The invention belongs to the field of signal processing, and particularly relates to a robust direction of arrival (DOA) estimation method based on sparse and low-rank recovery. According to the technical scheme, firstly, based on a low-rank matrix decomposition method, a received signal covariance matrix is modeled as the sum of a low-rank noise-free covariance matrix and a sparse noise covariance matrix; then the convex optimization problem about a signal and noise covariance matrix is constructed based on a low-rank recovery theory; then a convex model about the sampling covariance matrix estimation error is constructed, and a convex set explicitly includes the convex optimization problem; and finally, based on the obtained covariance matrixes, DOA estimation is achieved through a MVDRmethod. In addition, based on the statistical characteristic that the sampling covariance matrix estimation error submits to progressive normal distribution, an error parameter factor selection criterion is derived to reconstruct the covariance matrixes. Numerical simulation shows that under the limited sampling conditions, compared with traditional CBF and MVDR algorithms, a proposed algorithm ishigh in DOA estimation accuracy and robust in performance.

An encoding system is configured to allow data to be transmitted at one of two selectable bit-error-rate quality factors. The first bit-error-rate quality factor selection corresponds to the conventional ATSC FEC encoding systems, and the second bit-error-rate quality factor selection provides an ATSC-like FEC encoding scheme that substantially improves the bit-error-rate. The first quality factor selection effects a 2 / 3 trellis encoding, whereas the higher quality factor selection effects a 1 / 3 trellis encoding. Because the high-quality trellis encoding rate of 1 / 3 is half the lower-quality trellis encoding rate of 2 / 3, the data rate of this high-quality encoded bit-stream is half that of the conventional lower-quality encoded bit-stream. The 1 / 3 trellis encoding is effected using an ATSC-compatible encoding and a modified symbol mapping. The encoding scheme provides 2:1 data redundancy and the symbol mapping provides a maximum distance for the redundant encoding. By combining techniques that each decrease the likelihood of an uncorrectable error at the receiver, the substantial improvement in bit-error-rate can be achieved. At the receiver, a single trellis decoder with different metric tables is used to decode the two bit-streams, thereby providing substantial compatibility with ATSC-compatible receivers.

The invention relates to a stability evaluation method and apparatus for a landslide prediction model. The method comprises a factor selection step, a combination generation step, and a stability evaluation step of obtaining an ROC curve, based on different combinations, of the landslide prediction model by applying test data and evaluating the stability of the landslide prediction model according to the ROC curve. The method and the apparatus are used for evaluating the stability of the landslide prediction model, and a model with relatively high stability is used for performing landslide prediction, so that the landslide prediction accuracy can be improved; and the model with relatively high stability can be obtained through the method according to the landslide prediction model of any region.

The invention relates to a human face feature extracting method based on an active shape model and a POEM (patterns of oriented edge magnituedes) texture model in the complicated background and belongs to the technical field of model identification. The human face feature extracting method includes calibrating feature points of a training set; establishing an overall shape model for training samples; establishing a POEM texture histogram for each calibrated feature point; selecting initial human face shapes of a factor selection model according to the shape model; calculating the POEM histogram of each candidate feature point in a test image; calculating similarity of the candidate feature points and target points of the histogram by the mahalanobis distance function; performing iterative search matching by downloading initial human faces into the model; secondarily extracting local organs or human face outlines with poor extraction effect. By the human face feature extracting method, robustness of changes of complicated environments (such as posture, light and expression) is improved, high extraction accuracy is obtained, and the human face feature extracting method has good application prospect.

The invention provides an intelligent traffic signal lamp which comprises a traffic signal lamp and a prediction apparatus connected to the traffic signal lamp. The prediction apparatus comprises an acquisition module data, a data preprocessing module, a data classification module, a smooothness examination module, a correlation coefficient calculating module, a threshold setting module, a space-time correlation coefficient matrix generation module, a history correlation coefficient matrix generation module, a prediction factor selection module and a prediction model construction module. The prediction precision of the invention is high, and the constructed prediction model has pertinency.

The invention discloses a short-term traffic flow forecasting device, which comprises a data recovery module and a forecasting device, wherein the forecasting device comprises an acquisition module, a data pre-processing module, a data classification module, a stationarity test module, a correlation coefficient calculation module, a threshold setting module, a space-time correlation coefficient matrix generation module, a historical correlation coefficient matrix generation module, a forecasting factor selection module and a forecasting model construction module connected in sequence. The forecasting precision is high, and the constructed forecasting model is more targeted.

The invention provides a traffic accident data intelligent analysis and comprehensive application system. The system comprises a data docking module, a mining processing module, an interaction module,a map module and a data analysis module, the mining processing module drives data processing through a traffic accident data factor importance analysis model according to traffic accident data extracted by the data docking module, and the importance degree of attribute factor set elements is obtained; the data analysis module receives an attribute factor selection result of the interaction module, takes an attribute factor as a data analysis angle, and provides a targeted data analysis result through an accident data analysis mode; according to the system, importance analysis of attribute factors is carried out based on original traffic accident data, a missing data estimation strategy is configured, and the situation that the attribute factors provided by different data sources of a traffic accident are inconsistent can be effectively coped with; therefore, attribute factors with important traffic accident information are extracted from the sample data selected by the user, and quantitative importance indexes are output.

The invention provides a variable class remote sensing image segmentation method based on optimal fuzzy factor selection. The method comprises: reading a to-be-segmented remote sensing image; determining an optimal homogeneous region class number of the to-be-segmented remote sensing image; finding the homogeneous region class of each pixel spectrum measurement vector in the to-be-segmented remote sensing image through de-fuzzification, and obtaining a segmentation result of the to-be-segmented remote sensing image. The method adopts the partition entropy index as the index of a preferable fuzzy factor. When the fuzzy factor of the to-be-segmented remote sensing image is smaller than the optimal fuzzy factor, a PE index is larger. When the fuzzy factor is just equal to the optimal fuzzy factor, the PE index jumps to a smaller value, and the PE index becomes stable gradually when the number of fuzzy factors further increases. The corresponding minimum fuzzy factor when the PE index converges is selected to be the optimal fuzzy factor. The class number corresponding to the optimal fuzzy factor is the optimal class number, so that the class number of homogeneous regions in the remote sensing image is determined, and a better segmentation result can be obtained.

A POOL-mechanism-based cluster regulation algorithm for target tracking in WSN (Wireless Sensor Network) belongs to the technical field of wireless sensor networks and is used for target tracking in a wireless sensor network. In the traditional algorithm for target tracking, when the target moves slowly or is in the process of approximating stillness, a cluster head node consumes energy too quickly for serving as the cluster head for a long time, so that energy hole is caused easily. The algorithm provides a cluster head rotation mechanism based on a POOL mechanism; during the process of tracking a target with a variable motion in the wireless sensor network, a threshold value is set at the time when a tracking node serves as the cluster head, and when the set time is up, another node is selected from a POOL (a cluster head pool) to serve as a new cluster head according to the factors such as surplus energy, the RSSI (Received Signal Strength Indication) aparts from a target node, and the like. During the process of variable motion of the target, the structure of the POOL and the data update unceasingly. The mechanism ensures smooth transfer of the cluster head and can reduce energy consumption and prolong the service life of the network while ensuring the tracking accuracy.

A timing generator and methods thereof are provided. In a first example method, a timing control signal may be produced by generating a base clock signal and a higher delay resolution clock signal, a clock cycle of the higher delay resolution signal being less than a clock cycle of the base clock signal. A first control word output signal may be generated by synchronizing a control word with the base clock signal. A second control word output signal may be generated by synchronizing the first control word output signal with the higher delay resolution clock signal and generating at least one additional control word output signal based on the second control word output signal and the higher delay resolution clock signal, the first, second and at least one additional control word output signal each having different delay resolutions. In a second example method, a timing control signal may be produced by generating a plurality of control word output signals, each of the plurality of control word output signals having a different delay resolution and selecting one of the plurality of control word output signals based on a delay resolution of the selected control word output signal, the delay resolution of the selected control word output signal better suited for interaction with an external device than delay resolutions of other of the plurality of control word output signals. A timing generator may be configured to perform either of the above-described first and second example methods.

The invention discloses an automobile hub bearing fault feature extraction method based on optimal quality factor selection. Firstly, a vibration signal is collected; then resonance sparse decomposition parameters are initialized; by utilizing a progressive optimization algorithm, an optimal quality factor is obtained by taking an RSK index as a target function; and finally, envelope analysis is performed for a low-resonance component obtained by carrying out resonance sparse decomposition on the signal under the optimal quality factor to obtain an envelope spectrum, so that fault features canbe effectively extracted. According to the method, the problems of high randomness, uncertainty and difficulty in obtaining an ideal decomposition effect due to manual selection of the quality factorin a traditional resonance sparse decomposition method are avoided; the optimal quality factor can be selected in a self-adaptive mode; and the fault features of an automobile hub bearing under intermittent strong-interference noises can be effectively extracted.

The invention relates to a ship thrust distribution system and a ship thrust distribution method. The ship thrust distribution system comprises a parameter setting module, a limiting factor selectionmodule, a mathematical model establishment module, an optimization algorithm selection module, a distribution scheme generation module, a data storage module and an analogue simulation module. The ship thrust distribution method comprises the steps of: S1, setting a layout and parameters of a propeller; S2, setting limiting factors; S3, establishing a thrust distribution mathematical model; S4, selecting an optimization algorithm; S5, generating a thrust distribution scheme according to the thrust distribution model and an optimization algorithm; and S6, generating a plurality of thrust distribution schemes by adjusting the parameters. According to the ship thrust distribution system and the ship thrust distribution method, various thrust distribution schemes for different ships can be completed only by setting various parameters, and the different thrust distribution schemes can be simulated at the same time, so that a large amount of time for compiling simulation programs and analyzing simulation data is saved, and meanwhile, a large amount of cost can be reduced.

The invention provides a partial transmission sequence peak-to-average ratio suppression algorithm of boundless information suitable for a spare underwater acoustic OFDM communication system, in particular relates to a method for solving the problem that the PTS algorithm needs to transmit side information through the adoption of the inherent sparsity of an underwater acoustic channel and the side information blind detector based on the matching pursuit channel reconstruction technology. The pectinate pilot frequencies distributed at different positions are used for carrying weighting factor selection information, and the matching pursuit technology is used for reconstructing the received different pilot frequency domain responses, and the high-reliability weighting factor blind detection can be realized through the combination with the inherent sparsity of the shallow sea underwater acoustic channel, the reliability of the communication system is guaranteed. Since the matching pursuit channel reconstruction technology adopted in the invention needs less pectinate pilot frequency number and is free from uniform; compared with the traditional PTS algorithm, the improved PTS peak-to-average ratio suppression algorithm has better PAPR suppression performance; and meanwhile, the communication frequency band utilization rate and the communication efficiency are further improved; the algorithm is suitable for the underwater acoustic communication system in resource shortage.

The invention discloses a binary classification oriented factor screening method based on boosted regression trees. The method comprises the following steps that: (1) searching data, and establishinga target variable-predictive factor dataset; (2) on the basis of the target variable and all factors, utilizing the boosted regression trees to carry out modeling, and calculating and sorting factor importance; (3) carrying out correlation analysis on all factors, analyzing a Pearson correlation matrix, and carrying out screening; (4) on the basis of the target variable and the retained factor, utilizing the boosted regression trees to establish a new model, calculating a predictive deviation, calculating and sorting the factor importance, and removing the factor with the lowest importance until the amount of the retained factors is less than or equal to 2; and (5) comparing the predictive deviation of each boosted regression tree model in the (4), and taking all factors adopted by the boosted regression tree model with the smallest predictive deviation as an optimal factor combination. By use of the method, a quantitative factor selection system is established, results are reliable, and an application field is wide.

The invention discloses a vegetable harvesting cutter cutting factor selection testing platform. A vegetable fixing part is used for fixing vegetables on a testing operating table by virtue of a clamping bolt and rightly adjusting the positions of the vegetables; a vegetable clamping and extracting part is used for clamping the vegetables by adjusting the distance between adjustable clamping claws and adjusting a rotating angle of a clamping plate fixed shaft; a cutter harvesting cutting part is used for adjusting a slip cutting angle of a cutter and the vegetable cutting height by adjusting an angle between an adjustment angle adjusting block and a horizontal plane and adjusting the height from a vegetable fixing plate; the cutting speed of the cutter is controlled by adjusting the speed of stepping a synchronous belt on a linear sliding table; cutting force of the cutter is measured by using a force measuring sensor; and after being completely cut, the clamped vegetables are lifted by the vegetable clamping and extracting part. According to the vegetable harvesting cutter cutting factor selection test platform disclosed by the invention, comprehensive factors of vegetable harvesting cutting can be selected, optimal cutting factors can be determined, and the success rate and yield of a vegetable harvesting process can be increased; and the testing platform is easy to operate and strong in practicability.

The embodiment of the invention provides an indirect health factor selecting method for lithium battery capacity estimating. The method comprises the steps of calculating the relevancy between the battery capacity and a plurality of backup indirect health factors; selecting N backup indirect health factors with the maximum relevancy from the multiple backup indirect health factors as initial indirect health factors; adopting the battery capacity corresponding to the initial indirect health factors and the initial indirect health factors as input of a related vector machine model, and performing model training on the related vector machine model to obtain a target training model; judging whether the target training model meets the preset precision requirement or not; if the target trainingmodel meets the preset precision requirement, adopting the initial indirect health factors as target indirect health factors so as to be used for estimating the lithium battery capacity; if the targettraining model does not meet the preset precision requirement, updating N as N+1. The purpose of indirect health factor selection can be well achieved.

The invention provides an intelligent traffic signal lamp which comprises a traffic signal lamp and a prediction apparatus connected to the traffic signal lamp. The prediction apparatus comprises an acquisition module data, a data preprocessing module, a data classification module, a smooothness examination module, a correlation coefficient calculating module, a threshold setting module, a space-time correlation coefficient matrix generation module, a history correlation coefficient matrix generation module, a prediction factor selection module and a prediction model construction module. The prediction precision of the invention is high, and the constructed prediction model has pertinency.

Disclosed are a system and a method for selecting index Portfolio Managers / Products and active Portfolio Managers / Products for an investment portfolio. The invention described herein separates the performance impact of temporal market events from a Portfolio Manager's active security and / or factor selection skill. Furthermore, the invention described herein uses forecasting methods to improve the accuracy with which investors can select Portfolio Managers / Products that are likely to outperform their peers. In one embodiment, the method prepares data by calculating excess returns for each Portfolio Manager / Product within a universe of Portfolio Managers / Products using stock market indices and extracting Active Share from various datasets. Additionally, the method extracts raw factor data and generates composite indices for sectors. Several analytical inputs are then generated using edge measure and skill score measure. Each Portfolio Manager's / Product's rolling excess return is quartiled or segmented and a logistic regression model is calibrated to forecast performance.

The invention relates to a method for detecting spatial aggregation of urban water network leakage and identifying key influencing factors. The method comprises the following steps: 1, loading basic data information of urban water network leakage; 2, calculating the spatial aggregation of the urban water network, which comprises the following sub-steps: sub-step 1) determining a scanning mode, that is, how to define a position and a size of a scanning window; sub-step 2) calculating statistics; sub-step 3) carrying out significance analysis; sub-step 4) carrying out cluster analysis; 3, calculating the reliability score of the spatial aggregation of the urban water network; 4, analyzing that evolution of the spatial agglomeration of the urban water network; 5, establishing a prediction model of factors affect that spatial agglomeration of the urban water network based on the probabilistic neural network, which comprises the following sub-steps: sub-step 1) carrying out dependent variable selection; sub-step 2) carrying out factor selection; sub-step 3) carrying out probabilistic neural network modeling; 6, screening key influencing factors.

The invention provides a quantitative selection method and system for a smooth floating datum in seismic data processing. The method comprises the steps that smooth floating datum target functional establishing and minimizing are carried out; a maximum smoothing factor is determined; a minimum smoothing factor is determined; and an arbitrary value between the maximum smoothing factor and the minimum smoothing factor is selected to acquire the smooth floating datum. According to the invention, a quantitative calculation formula for constructing the smooth floating datum is provided, which makesthe construction of the smooth floating datum more scientific; and in addition, the smoothing factor selection basis and the determination selection range of the smooth floating datum are provided, which makes the selection of the smooth floating datum more reasonable.

The invention provides a multi-screen display device capable of carrying out traffic flow prediction. The multi-screen display device capable of carrying out traffic flow prediction comprises a multi-screen display device and a prediction device connected with the multi-screen display device, wherein the prediction device comprises an acquisition module, a data pre-processing module, a data classification module, a smooothness detection module, a correlation coefficient calculation module, a threshold setting module, a space-time correlation coefficient matrix generation module, a historical correlation coefficient matrix generation module, a prediction factor selection module and a prediction model construction module which are sequentially connected. The multi-screen display device capable of carrying out traffic flow prediction is advantaged in that prediction precision is relatively high, and a constructed prediction model is with more pertinency.

An optimal route searching system of the present invention comprises a searching system and a prediction device connected with the searching system, and the prediction device comprises an acquisition module, a data pre-processing module, a data classification module, a stability examination module, a correlation coefficient calculation module, a threshold setting module, a space-time correlation coefficient matrix generation module, a historical correlation coefficient matrix generation module, a prediction factor selection module and a prediction model construction model which are connected orderly. The optimal route searching system of the present invention is higher in prediction precision, and a constructed prediction model is more targeted.

Disclosed are a system and a method for selecting index Portfolio Managers / Products and active Portfolio Managers / Products for an investment portfolio. The invention separates the performance impact of temporal market events from a Portfolio Manager's active security and / or factor selection skill. The method includes preparing data by calculating excess returns for Portfolio Managers / Products using stock market indices, extracting Active Share, and extracting raw factor data and generating composite indices for sectors. Using the skill metrics, Active Shares, and manager 36-month return, a cross sectional rolling regression model with rolling one-month window is calibrated to forecast the probability of outperforming a benchmark over the subsequent 36-month period. To determine the efficacy of each of the forecast models, an analysis is performed to determine the overall accuracy for each one. P-values are used to measure significance of the independent variables. Accuracy is measured by comparing forecasts with managers' actual excess returns.