52results about How to "Realize deep mining" patented technology

The present invention discloses a series-wound long short-term memory recurrent neural network-based heating load prediction method. The method comprises the steps of constructing a sample data set based on temperature, climate and heat supply data during a given period of time, and respectively subjecting the input data and the output data of the sample data set to standardized treatment; dividing the input data into two portions, respectively inputting the two portions into two independent long short-term memory recurrent neural networks to merge the two portions of the input data, inputting the output data to a long short-term memory recurrent neural network at a next layer, and finally inputting the data into two full connection layers; training a constructed series-wound long short-term memory recurrent neural network, and optimizing the network by adopting the parameter optimization-based adaptive torque estimation algorithm; inputting to-be-predicted data into the series-wound long short-term memory recurrent neural network, calculating and obtaining a heating load prediction result. The method of the invention can effectively discriminate input data, and accelerate the learning speed. Therefore, the learning efficiency is improved and the prediction accuracy is improved.

The invention provides a small area-level ultra-short-term load prediction and visualization method based on a Long Short-Term Memory (LSTM) network, the method comprising: step 1: an input and outputvariable of the model is determined; step 2: The input and output data sets are separately preprocessed; step 3: the depth LSTM load prediction model is constructed and the random search method is used to find the appropriate hyperparameters until the test set prediction error is minimized. step 4: The t-SNE visualization technology is used to visually characterize the network hidden layer vector, and the correlation coefficient heat map is formed according to the hidden layer vector to perform correlation quantitative analysis, thereby reflecting the network's ability to extract feature datafrom the input data. The method aims to utilize the feature extraction capabilities of the deep learning model and the LSTM temporal correlation learning capabilities to achieve higher prediction accuracy than the machine learning model.

The invention provides an infectious disease prediction method and apparatus, an electronic device and a computer readable medium, and relates to the technical field of computer application. The infectious disease prediction method includes the steps: obtaining historical occurrence data of infectious diseases; extracting spatial features of historical occurrence data through a convolutional neural network; extracting first time sequence features of the historical occurrence data through a cyclic neural network; and fusing the spatial features and the first time sequence features to obtain temporal and spatial features of the historical occurrence data; and performing risk prediction on infectious diseases based on the temporal and spatial features. The infectious disease prediction methodand apparatus utilize the prediction network model composed of the convolutional neural network and the cyclic neural network to deeply mine the temporal and spatial features of the infectious disease data, thereby improving the accuracy of infectious disease prediction.

The invention provides an audio watermark-based information interaction method and a device. The audio watermark-based information interaction method comprises steps: sound signals played in the surrounding environment are acquired; an audio watermark inserted in the sound signals is extracted, wherein the audio watermark is generated and embedded according to an information identifier corresponding to real-time information which needs to be provided for a user by a sound signal playing source; the information identifier is analyzed in the audio watermark, and a corresponding information interaction instruction is obtained according to information identifier matching; and according to the information interaction instruction, information interaction with an information interaction server specified by the sound signal playing source is carried out. By applying the scheme of the invention, an additional information interaction means is provided for traditional media, deep utilization can be carried out on contents of the traditional media, the business outlook is good, and more values can be provided for a media provider.

The invention relates to a medical data joint learning system and method based on trusted computing and privacy protection. A joint learning center control layer receives non-sensitive meta-information uploaded by a data contributor through a data contributor management layer of a data node where the joint learning center control layer is located for filing, and original data is locally registered, stored and subjected to isolation calculation; and the joint learning center control layer processes a joint learning request initiated by the data miner through the data miner interaction layer, summarizes insensitive intermediate results obtained by performing local isolation calculation on the data nodes based on the original data in a safe calculation area, and returns a final joint learningresult to the data miner interaction layer. The invention provides a whole set of service system based on medical big data security sharing, trusted computing, deep mining, authority authentication and multi-platform joint learning, and solves the problems of scattered, single and incomplete medical data privacy protection and data mining at the present stage.

The invention provides a thermal power plant condenser vacuum degree prediction method based on multilayer LSTM. Firstly, a sample data set is constructed, input data and output data in a sample are subjected to standardization processing respectively, then the input data sequentially enter a multi-layer long-short-term memory neural network, and model training is performed by using an adaptive moment estimation algorithm; and finally, to-be-predicted data is input into the trained prediction model, and the vacuum degree of the condenser is predicted. According to the method, the long-term andshort-term memory neural network structure is applied to condenser vacuum degree analysis, deep mining of big data is achieved, the accuracy and speed of data prediction are improved, and use of system resources is optimized. According to the two-layer LSTM structure designed by the invention, the model depth can be increased, the model capacity can be increased, the prediction precision can be improved, and meanwhile, the adaptive moment estimation algorithm is used for optimizing the traditional gradient descent algorithm, so that the problem of gradient explosion is avoided.

The invention discloses a distribution network layering load forecasting method based on topology. A distribution network is layered according to a topology connection relation, forecasting is conducted on the layers respectively, a total forecasting result is obtained by summing in an optimizing mode, summarizing load data of a 10 kV outgoing line switch are read, and contrastive analysis is conducted on the data and the forecasting result. Forecasting accuracy is improved obviously, an intelligent distribution network system is utilized effectively, deep mining based on big data is achieved, the immunity to the data is improved, and lean management of the distribution network is achieved easily.

The embodiment of the invention provides a user portrait analysis method based on e-commerce big data and an artificial intelligence platform. The method comprises the steps of obtaining browsing behavior information of a video live broadcast terminal of a live broadcast audience account for live broadcast commodity recommendation information of the live broadcast audience account; performing deepanalysis to obtain a current positive browsing relationship feature and a current negative browsing relationship feature; and obtaining browsing tendency information after comparison, determining browsing behavior label information of the anchor audience account based on the browsing tendency information, and then analyzing the browsing behavior label information of the anchor audience account according to a preset artificial intelligence model to generate a user portrait of the anchor audience account. Thus, deep mining of further browsing behaviors of audiences on the live broadcast commodity recommendation information can be realized, so that the browsing tendency of the audiences is effectively mined, the deep dimension of the analysis process of the user portraits of the audiences isexpanded, and subsequent information pushing and audience experience optimization are facilitated.

The invention relates to an integrated platform for creative design based on the recycling of cultural relics elements of Changsha kiln, belonging to the application field of cultural relics creativedesign based on artificial intelligence. The platform integrates the semantics of Changsha kiln cultural relics and graphics and image features, integrates the image style transfer technology, and provides creative designers with the full-process creative design service support which starts from retrieving accurate, detailed and accurate cultural relics materials, integrates the characteristics ofother craft products and automatically forms a personalized material library, and automatically synthesizes the pattern of the target artifacts. The platfrom realizes the deep excavation of the cultural relics' internal cultural resources and external artistic resources and the automatic generation of creative materials, enriches the creativity, saves the human resources, design and development time and cost.

The invention discloses an information mining method, an electronic device and a readable storage medium. The method comprises the steps of obtaining a specific type of information from a predetermined data source in real time or in regular time; conducting word-segmentation processing on the obtained information, and conducting part-of-speech tagging on each word segmentation corresponding to each information; according to the word-segmentation sequence and part of speech of each word segmentation corresponding to the information, constructing each word segmentation corresponding to the information into a word-segmentation tree of a preset structure; after the word-segmentation structure of the preset structure corresponding to the information is constructed, according to the word-segmentation structure of the preset structure corresponding to the information, parsing out core opinion information corresponding to the information. Through the adoption of the method, deep mining of the information is achieved, and the core opinion information in the information is obtained accurately.

The invention discloses a deep learning-based river network dense drainage basin water regime trend prediction method and application thereof, and the method comprises the following steps: collecting daily scale rainfall and water level time sequence point data of all rainfall stations and river channel water level stations in a whole drainage basin range, and longitude and latitude information of each observation station, and recompling the above; converting the reorganized rainfall and water level time sequence point data into spatio-temporal data through an inverse distance weight method and a contour surface diagram rendering method; constructing and training a watershed high-resolution water level forecasting model, and respectively inputting rainfall and water level spatio-temporal data tensors; wherein the watershed high-resolution water level forecasting model comprises a preliminary spatial feature extraction module, a spatial-temporal feature extraction module, a stacking module, a splicing module and a feature reduction module; and the model outputs a water level contour surface diagram of the whole watershed in the prediction period to obtain a watershed water regimen change trend. The runoff production and confluence mechanism of the drainage basin can be fully mined from historical hydrological data, and the space-time water regimen change of the whole drainage basin can be accurately forecasted.

The invention discloses a data acquisition system and method based on a scrapy crawler framework. The data acquisition system comprises a crawler queue module and a crawler execution module, the crawler queue module comprises a crawler seed queue, a crawler seed processing unit and a crawler task queue; wherein the crawler seed queue is used for storing crawler tasks; the crawler seed processing unit is used for carrying out deduplication screening processing on crawler tasks in the crawler seed queue and storing the crawler tasks subjected to deduplication screening into a crawler task queue;the crawler execution module comprises a webpage downloading unit and a URL mining unit; the webpage downloading unit is used for reading a crawler task needing to be executed currently from the crawler task queue and downloading a webpage based on the read crawler task; the URL mining unit is used for extracting a new URL link from the downloaded webpage to serve as a new crawler task to be stored in a crawler seed queue; deep mining of website domain names in a specific field is realized, and the crawling range of the system is improved.

The invention belongs to the technical field of integrated energy dispatching, and discloses a source-load collaborative carbon reduction integrated energy system optimal distribution robust economic dispatching method, which comprises the following steps of: constructing an integrated energy system carbon emission model; establishing a source side joint operation framework on the source side; setting a price type comprehensive demand response mechanism on a load side; constructing a demand response model; comprehensively considering different energy utilization ratio conditions of a load side, setting an energy utilization ratio coefficient, and constructing a satisfaction evaluation model; therefore, an integrated energy system scheduling mode based on a source-load collaborative carbon reduction mechanism is established; based on a distribution robust optimization theory of an imprecise probability opportunity constraint, a constraint relation between robustness and economy of a scheduling scheme is fully considered, an optimal distribution robust economic scheduling model of the integrated energy system is constructed, and a compromise solution is solved, so that subjective limitation of conventional distribution robust optimization is eliminated, and the optimal distribution robust economic scheduling model of the integrated energy system is obtained. And finally, comprehensive energy system optimization scheduling capable of balancing robustness and economical efficiency is realized.

The invention relates to a solid rocket engine mapping design method and device, and equipment. The method comprises the steps: obtaining a target thrust curve and an engine original parameter of a solid rocket engine ; performing geometric scaling on the engine original parameter; according to the engine parameter subjected to the geometric scaling, calculating the shell area, the heat insulation layer area, the charge combustion area and the thickness; calculating the nozzle throat area according to the engine parameter and the charge combustion area; determining the engine combustion speed through an optimization algorithm, and determining the throat diameter is determined according to the nozzle throat area; generating a design thrust curve according to the engine parameter, the thickness, the engine combustion speed and the nozzle throat area; calculating the root-mean-square error of the design thrust curve and the target thrust curve, and adjusting the engine combustion speed according to the root-mean-square error by using an optimization algorithm; and outputting the corresponding overall mapping design parameter of the engine when the root-mean-square error is minimum. By using the technical scheme, the technical effect of greatly improving the design efficiency is achieved.

The invention discloses a model training method and device, equipment and a storage medium. The method comprises the steps of obtaining a first travel sample; obtaining a feature extraction model; processing the first travel sample through a feature extraction model to obtain a travel feature vector, and processing the travel feature vector through a first classification model to obtain first prediction information; and respectively updating the feature extraction model and the first classification model according to the first prediction information and the mark information corresponding to the first travel sample to obtain a first risk control model. According to the model training method and device, the equipment and the storage medium, the accuracy of the first risk control model obtained through training can be improved.

The invention discloses a physical examination diagnosis and treatment system and method for improving enterprise innovation ability, which belong to the technical field of enterprise diagnosis and treatment systems. The physical examination diagnosis and treatment system comprises a physical examination diagnosis and treatment module, the physical examination diagnosis and treatment module is connected with a data transmission module through the Internet, and the data transmission module is connected with an innovation processing unit through the Internet; the innovation processing unit is connected with a central processing unit through the Internet, the central processing unit is connected with a data comparison module through the Internet, the data comparison module is electrically connected with an index evaluation module through the Internet, and the index evaluation module is connected with a physical examination report management module through the Internet. The system can achieve the deep mining of the enterprise innovation capability, can plan the innovation direction according to the actual conditions of an enterprise, is high in work efficiency, and is high in practicality.

A link prediction method base on common neighbor node resource allocation and naive Bayesian is disclosed, By establishing network model G, any two unconnected nodes x and y and the common neighbor nodes of node x and y in network G are selected to calculate the distribution value of node x and node y under the action of the common neighbor nodes. Secondly, the connection attribute function of thecommon neighbor node in step S1 is obtained by using the naive Bayesian method, and the role difference of the common neighbor node is distinguished by the connection attribute function; Finally, thefinal similarity value of any two unconnected pairs of nodes in the network G is calculated by combining the assignment value between the pairs of nodes to be predicted and the connection attribute function of the common neighboring nodes, and the network link prediction is carried out according to the final similarity value of the pairs of nodes to be predicted. The invention utilizes the naiveBayesian method to supplement the attribute differences between different nodes, so that the link prediction accuracy can be effectively improved.