1962 results about "Outcome predictor" patented technology

A method comprises receiving a network of a plurality of nodes and a plurality of edges, each of the nodes comprising members representative of at least one subset of training data points, each of the edges connecting nodes that share at least one data point, grouping the data points into a plurality of groups, each data point being a member of at least one group, creating a first transformation data set, the first transformation data set including the training data set as well as a plurality of feature subsets associated with at least one group, values of a particular data point for a particular feature subset for a particular group being based on values of the particular data point if the particular data point is a member of the particular group, and applying a machine learning model to the first transformation data set to generate a prediction model.

The invention discloses a day-dimension regional traffic index prediction method considering influences of multiple factors. The method comprises the steps that regions are divided and aggregated; regional traffic index original data preprocessing is carried out; the influences of multiple factors are considered, and regional traffic index prediction under the day dimension is carried out. According to the specific technical scheme of the method, on the basis of traffic cell division, traffic cells with the same aggregation property are aggregated, and regional traffic indexes are calculated;on the basis of road network operation early warning requirements, a prediction time period and a prediction cycle are determined; regional traffic data is extracted, made up for and removed, and preprocessing such as comprehensive building of a historical data factor attribute set from different angles is conducted on the data; on the basis of a decision tree theory, regional road network operation congestion state prediction is carried out; a final prediction result of the regional traffic indexes is determined by means of the square euclidean distance. By means of the method, on the one hand, monitoring and application of the urban road network operation state is deepened, and on the other hand, technical support is provided for early warning and forecasting work of the road network operation state.

The invention relates to a short-term load prediction method based on a clustering and sliding window. The method comprises the following steps of: preprocessing electric power load data; clustering historical data of a prediction user by utilizing a clustering algorithm, and adjusting clustering parameters; selecting k data from near to far of the prediction time in a category, containing most data, in clustering results to form a sliding window k; predicting the k selected data by utilizing a combination model based on the sliding window, and acquiring a primary prediction result; and correcting the primary prediction result of the combination model according to meteorological factors to obtain a final load prediction result. Compared with the prior art, the method has the advantages of high prediction precision, good adaptability and the like.

Systems and methods are provided that predict the accuracy of expert forecasts based on a corpus of prior expert forecasts. Expert forecasts (also referred to as opinions) are analyzed and processed to determine an opinion and related positions expressed by the expert. Each opinion is combined with relevant industry information to create a structured model for the opinion. A plurality of structured models for a given expert are then analyzed along with information related to the actual outcome of prior predictions to create a decision model for the particular expert. A new opinion issued by an expert is then analyzed using the decision model for that expert to predict the accuracy of the new opinion from the expert. Recommended actions in accordance with or against the new expert opinion may then be provided.

The invention regards a method for assisting a driver of a vehicle by computationally predicting a future movement behaviour of a target vehicle. The method comprises steps of acquiring data describing a traffic environment of a host vehicle, computing a plurality of future movement behaviour alternatives for a target vehicle, by predicting movement behaviour of the target vehicle based on the data describing the traffic environment. The prediction step calculates a probability that the target object will execute a movement behaviour and information relating to future behaviour of the target vehicle is output. The method estimates a global scene context and adapts the predicted behaviour based on the estimated global scene context. The adaptation of the behaviour prediction is performed by at least one of adapting indicator values in situation models employed by a context-based prediction, by adapting a result of the context-based prediction and adapting the overall prediction result.

The invention provides a short-term traffic flow prediction method based on a nerve network combination model. The method is used to construct a counterpropagation nerve network combination prediction model and the short-term traffic flow prediction method is provided based on the model. Aiming at a characteristic of a traffic flow, a fuzzy C mean value clustering algorithm is used to cluster the traffic flow. For bunch generated through clustering, a counterpropagation nerve network prediction model is constructed. According to grade of membership, a weighted sum of prediction model prediction results is calculated and is taken as a final prediction result. In order to increase prediction precision, a taguchi method is used to carry out test designing so as to test influences of different structure parameters on prediction model prediction precision, and an optimum structure parameter is used as an initial structure of the prediction model. By using the method in the invention, the prediction precision of the short-term traffic flow can be effectively increased, an influence of a noise on the prediction precision in training data is reduced and operation time is reasonable.

The invention discloses a target detection method based on full-automatic learning. The method comprises the following steps: carrying out the training of a model through employing a small-scale manual labeling data set after preprocessing and a deep neural network, carrying out the fine adjustment of the model trained through an Imagenet data set, and obtaining a deep model; carrying out reasoning prediction on the pseudo-labeled part of the original large-scale image data set by utilizing the deep model, removing repeated prediction of the same target after carrying out non-maximum suppression, and respectively storing bounding boxes and confidence of prediction results according to categories. Through self-supervised pseudo labeling and active learning sample selection, the informationentropy and the divergence degree predicted by the deep neural network are learned in a combined manner, unlabeled samples are sorted according to the weight, and pseudo labels are allocated to high-confidence samples ranked at the top. The objective of the invention is to solve the problems of too high labor cost of common target detection and labeling and poor mobility and adaptability of a training model in an actual scene.

A system and method which uses statistical analysis of college football games to predict an outcome is comprised of a unique combination of objective, quantitative factors. The system and method accounts for when in the college football season the contest takes place and uses varying sets of objective parameters based on the season game number. The system and method may be implemented on a variety of processing platforms which allows for rapid analysis of multiple contests.

The invention discloses a small-scale air quality index prediction method and system for a city, which firstly divides a city area into a plurality of to-be-predicted locations in a grid form; and then acquires historical data related to each model, and based on historical data: establishing corresponding correspondences The current time prediction and the time prediction model predicted at each moment in the future time, establish a spatial prediction model for air quality prediction at the specified coordinates, and establish a dynamic prediction model that characterizes the relationship between traffic data and geographic interest point data and air quality index, an indoor and outdoor prediction model that characterizes the relationship between the indoor air quality index and the outdoor air quality index; when performing the prediction, the established time prediction model, the spatial prediction model, and the dynamic prediction are performed for any real-time moments to be predicted. The model and the indoor and outdoor prediction models are cooperatively trained to fuse the prediction results of all the models, that is, the predicted values of the air quality index at each moment in the respective current and future time periods of each to-be-predicted location.

The invention discloses a support vector machine-based wind electric powder prediction device and a support vector machine-based wind electric powder prediction method, and belongs to the technical field of wind electric power prediction. The device comprises a microprocessor, a sensor, an external expanding memory, a data acquisition module, an energy management module (EMS) and a communication interface. The prediction method applying the device comprises the following steps of: 1, normalizing historical data; 2, acquiring meteorological information of a prediction day by using the data acquisition module to form a training sample set; 3, obtaining an optimal clustering centre matrix which meets accuracy requirement; 4, classifying the training sample set according to distances of clustering centers; and 5, obtaining the wind electric power prediction result. The device and the method have the advantages that: the induction principle of structural risk minimization is realized, the training is equivalent to that the quadratic programming problem of linear constraint is solved; and the device and the method have unique solution.

The invention discloses a small molecule drug virtual screening method based on deep migration learning and application thereof. A source domain is used as an input to be trained, converged and derived to obtain a weight matrix; a target domain is input into an improvement tool to serve as the initialization weight of the target domain; fine adjustment, training and convergence are conducted on the initialization weight and data in the target domain sequentially; a biological activity value of interaction of a lead compound and a drug target in the target domain is predicted, a target domain molecular fingerprint and a predicted value are obtained, and an evaluation index root mean square error and a correlation coefficient of a predicted result are output; the target domain is subjected to fine adjustment by repeating above steps, and the weight matrix of the source domain helps the target domain build a model. According to the small molecule drug virtual screening method and the application thereof, the effective virtual screening model can still be obtained under the condition that the information of a known active ligand sample is insufficient, and does not need to rely on a large number of data samples.

The invention discloses a personalized traffic accident risk prediction recommendation method based on depth learning, which comprises the following steps: dividing a city into grid regions; calculating traffic accident data, traffic flow data and weather characteristic data of each grid region and each period; using the depth learning method to train the model, the traffic accident risk prediction model being obtained. According to the traffic accident data, traffic flow data and weather characteristics data input at the present time, the traffic accident risk prediction model is used to calculate the traffic accident risk prediction at the next time. The invention utilizes the depth learning method to learn the non-linear, high-dimensional and complex correlation relationship between thetraffic accident influence factor and the traffic accident, predicts the traffic accident risk at the city level, and improves the accuracy of the prediction result.

The invention relates to a collaborative anti-cancer pharmaceutical combination prediction method and a pharmaceutical composition. The collaborative anti-cancer pharmaceutical combination prediction method comprises the following steps: 1) data collection: according to different disease treatment effects of a pharmaceutical combination, classifying and obtaining a known collaborative anti-cancer pharmaceutical combination and a corresponding target; 2) model establishment: for the known collaborative anti-cancer pharmaceutical combination and an unknown pharmaceutical combination, calculating a characteristic of the collaborative anti-cancer pharmaceutical combination, and establishing a collaborative anti-cancer pharmaceutical combined prediction model; and and 3) result filtration: expressing spectrum information with the pharmacy, exploring and inducing the characteristic of the known collaborative anti-cancer pharmaceutical combination, and conducting screening with prediction results of the step 2). An anti-breast cancer pharmaceutical combination and an anti-lung cancer pharmaceutical combination can be acquired on the base of the collaborative anti-cancer pharmaceutical combination prediction method. Compared to the prior art, according to the invention, the collaborative anti-cancer pharmaceutical combination prediction method comprehensively uses various characteristics of the pharmaceutical combination, is designed dexterously, predicts accurately, has an important practical application and is suitable for large-scale popularization.

The invention discloses a prediction method for an unbalanced data set based on isolated forest learning. The prediction method comprises the following steps: receiving a prediction request; collecting data, and defining features and labels in the data set and the number of minority class samples and majority class samples; converting a non-numerical feature column and a label column in the data set into classification numerical values; synthesizing minority class samples by using a majority class weighted minority class oversampling technology to form a balance data set; performing abnormal point identification and removal on the balance data set by using an isolated forest algorithm; then performing data standardization, and dividing a training set and a test set; constructing and training a support vector machine classifier model by using the training set; adjusting hyper-parameters of the support vector machine classifier model through a genetic algorithm, and obtaining a prediction model after training is completed; and inputting the test set into the prediction model to obtain a prediction result. The prediction method for the unbalanced data set based on isolated forest learning has the characteristics of stable prediction result and high prediction precision.

The invention relates to a data processing method and discloses a method for processing abnormal data of a real-time data acquisition system in real time. The method comprises the steps of (1) initializing sample data and selecting an even number of normally operating sample data; (2) adopting 1/2 of the sample data to act as the moving step by using a single exponential smoothing method, and predicting the latter half part of the sample data by using a single exponential smoothing recurrence method; (3) the residual of a prediction result is calculated according to a prediction value and a measured value of the latter half part; (4) carrying out anomaly analysis on the residual sequence according to a Pauta criterion to confirm whether the measured value is abnormal data or not; and (5) replacing the measured value with the prediction value if the measured value is abnormal data. The method disclosed by the invention is mainly advantageous in that a prediction algorithm coefficient is adjusted in an adaptive mode, the error is analyzed by adopting a mobile exponential smoothing method, and the anomaly judging method better conforms to use conditions of the Pauta criterion, thereby improving the accuracy in judgment for the abnormal data, and preventing false judgment and missing judgment to a certain degree.

The invention discloses a rolling bearing residual life prediction method based on a deep generative adversarial network. The method comprises the following steps: collecting an original vibration signal of a rolling bearing; acquiring characteristic parameters of the original vibration signal; dividing the feature parameters into a training set and a prediction set; sending the training set intoa generator long short-term memory network for training; predicting the degradation process of the rolling bearing, and generating a prediction result; building an automatic encoder model as a discriminator, and discriminating whether a prediction result is from real historical data or not; enabling a generator long short-term memory network and a discriminator automatic encoder to carry out adversarial training to seek an optimal solution; and outputting a rolling bearing residual life prediction result. According to the method, the degradation process of the rolling bearing is predicted through long-term and short-term memory network learning, the prediction result of the long-term and short-term memory network is judged through the automatic encoder, the two methods conduct adversariallearning till the precision requirement is met, the prediction error superposition problem of a traditional method is reduced, and the prediction accuracy is improved.

The invention discloses a method for scale prediction of a regional distributed type comprehensive energy-supply system. A scenario analysis method, a factor correlation analysis method and a market predicting model method are selected for conducting comprehensive prediction at different angles; prediction models are respectively established combined with the three prediction methods and prediction condition setting and prediction parameter setting are conducted on the three modules respectively; parameters and conditions are introduced into the prediction modules to obtain prediction results; the prediction results of the three methods are processed through an equal-weight allocation method, the stability and the practicability of the prediction results can be improved, and therefore a relatively-reasonable scale prediction value of the regional distributed type comprehensive energy-supply system is obtained. The method is simple in principle, diversified in prediction angle, good in prediction accuracy, capable of accurately predicting the development scale of a future distribution type comprehensive energy-supply system of each region and directing development and construction of the regional distributed type comprehensive energy-supply system, and high in practicability and popularization.

The present invention provides Systems and Methods for Creating an Optimal Prediction Model and Obtaining Optimal Prediction Results Based on Machine Learning. In the method for creating an optimal prediction model, the steps are first to input a plural training data and at least one of machine learning algorithms, then convert the training data into a relay format. The method is further to select the automated predictive features, optimize the machine learning algorithm parameter, and then optimize the iterative prediction model. After that, a prediction model and an accuracy assessment data are outputted. In the process of obtaining the prediction result, the data to be predicted is converted into a relay format, and an automated program is used for iterative prediction to generate and output the prediction result and accuracy evaluation data.

The invention belongs to the technical field of short term power load forecasting, and discloses a power load forecasting method based on improved exponential smoothing and a gray model. The method includes the following steps: inputting original power load real-time data, and conducting a single exponential smoothing on the original power load real-time data, weakening the randomness of the original power load real-time data, such that the original power load real-time data approaches exponential development trend; predicting a smoothed sequence by using a gray forecasting model which optimizes background value; conducting inverse exponential smoothing on the forecasting result and returning the result to original power load data and a forecasting value at a next forecasting moment; determining whether the result reaches the requirements of knitting fitting errors, and outputting a forecasting result. According to the invention, the method expands the application range of the gray forecasting model, shortens search intervals, has higher forecasting reliability as high as 97%, can the meet requirements for maintaining the average error of short term power load forecasting at approximately 3% so as to address the problem of short term power load forecasting in future development of intelligent power grids.

The embodiment of the invention provides a behavior prediction model training method. The method comprises the steps: firstly determining a plurality of sample users for a target object, enabling anyfirst sample user to correspond to a first sample hard tag, and indicating whether a specific behavior is carried out on the target object or not; performing graph embedding processing on the user-object bipartite graph based on the trained graph neural network to determine an embedding vector set; determining a sample user feature vector corresponding to the first sample user and a target objectfeature vector corresponding to the target object, determining a specific behavior probability that the first sample user makes a specific behavior to the target object as a first sample soft label, and inputting the sample user feature vector into a first behavior prediction model to obtain a behavior prediction result; and training the first behavior prediction model by using a first loss item determined based on the prediction result and the first sample hard tag and a second loss item determined based on the prediction result and the first sample soft tag.

The embodiment of the invention provides a pneumonia lesion segmentation method and device, and solves the problems of low accuracy and low efficiency of an existing pneumonia lesion segmentation mode. The pneumonia lesion segmentation method comprises the following steps: predicting a lesion area on medical image data of a positive level based on an image semantic segmentation model; counting thelesion area of each parallel layer, and calculating the lesion volume by combining the lesion area of each parallel layer, wherein the image semantic segmentation model is established through the following training steps: inputting all or part of marked sample data into a focus segmentation model to obtain a prediction result output by the focus segmentation model; based on a focus detection frame predicted by a focus detection model and a lung region predicted by a lung lobe lung segment segmentation model, screening out a low-level false positive region from a prediction result to obtain afalse label of sample data, and adding unmarked sample data; and rechecking the pseudo label, and marking the marked sample data to update the marked sample data.

The invention discloses a building foundation settlement monitoring and early warning method and a system, and relates to the field of foundation settlement monitoring. The method comprises the following steps: building a target building three-dimensional model; obtaining coordinate information of a sensor node through the target building three-dimensional model, obtaining displacement information of the target building according to the coordinate information change of the sensor node, and performing comparative analysis on the displacement information and historical monitoring data to generate a displacement deviation rate; whether the displacement deviation ratio is larger than a preset threshold value or not is judged, and if yes, foundation settlement alarm information is generated; and generating a target building settlement curve according to the monitoring data time sequence, establishing a target building settlement prediction model, performing rehearsal by using a target building three-dimensional model according to a prediction result, and generating a target building foundation settlement reinforcement scheme according to the prediction result. According to the invention, real-time monitoring of the settlement of the building foundation is realized, and more accurate data support and danger prediction are provided for monitoring of the settlement of the building foundation.

The invention relates to an elder falling detection method and system based on multiple classifier integration. The method includes acquiring information from sensors; performing falling direction, pressure, sound and misjudgment predictions; establishing falling characteristic vectors composed of the four prediction results; calling classifiers to implement integral falling prediction; outputting a falling judgment result. The system is characterized by comprising a sensor information acquisition module, a falling direction prediction module, a falling pressure prediction module, a falling sound prediction module, a falling misjudgment prediction module, a falling characteristic vector establishing module, an integral fall predicting module, an integral falling prediction model learning module, a falling early warning module and a falling file management module. The method and system has the advantages of high falling detection accuracy, low cost and convenience for carrying, and the alarming is performed timely when falling.