91 results about "Model interpretation" patented technology

A computer program based method is described for creating models using gestures. On an input device, such as an electronic whiteboard, a user draws a gesture which is recognized by a computer program and interpreted relative to a predetermined meta-model. Based on the interpretation, an algorithm is assigned to the gesture drawn by the user. The executed algorithm may, for example, consist in creating a new model element, modifying an existing model element, or deleting an existing model element.

A model-driven approach to business performance management (BPM) uses a hybrid compilation-interpretation approach to map an observation model to a runtime executable. The data aspect of the observation model is first extracted and refactored to facilitate runtime access. Next, the operational aspect of the model, such as logic for metric computation and situation detection, is compiled into code. Finally, a runtime engine interprets the refactored model and dynamically loads the generated code, according to the meta-model.

A computer program based method is described for creating models using gestures. On an input device, such as an electronic whiteboard, a user draws a gesture which is recognised by a computer program and interpreted relative to a predetermined meta-model. Based on the interpretation, an algorithm is assigned to the gesture drawn by the user. The executed algorithm may, for example, consist in creating a new model element, modifying an existing model element, or deleting an existing model element.

A system, method, and program product for interpreting, optimizing, and customizing data mining models through the use of statistical techniques that utilize diagnostic measures and statistical significance testing. A data processing system is disclosed that includes a data mining system for mining data from a data warehouse in accordance with a data model, wherein the data model defines how data groups can be partitioned; and a data group analysis system that calculates a set of diagnostic measures and performs statistical significance tests for a defined data group.

A computing system, comprising: an I / O platform for interfacing with a user; and a processing entity configured to implement a dialog with the user via the I / O platform. The processing entity is further configured for: identifying a grammar template and an instantiation context associated with a current point in the dialog; causing creation of an instantiated grammar model from the grammar template and the instantiation context; storing the instantiated grammar model in a memory; and interpreting user input received via the I / O platform in accordance with the instantiated grammar model. Also, a grammar authoring environment supporting a variety of grammar development tools is disclosed.

The invention discloses a system for developing, generating and managing large-data analysis model business. The system comprises a user terminal, a server terminal and a cloud calculating and storing platform. The user terminal comprises a user-model designing and generating subsystem and a local model database. The server terminal comprises a tenant managing subsystem, a model auditing and trading subsystem, a model explaining and compiling subsystem, a model executing and dispatching subsystem, a model metadata managing subsystem, a calculated-result showing subsystem, a model trading database, a metadata base and a model API database. According to a frame, an efficient large-data operation-and-maintenance mode with centralized-distributed controllable, manageable and tradable calculating resources, model resources and storage resources is adopted, and six functions such as data-distributed data storing and calculating platform, model designing and generating, model managing and trading, model-quality auditing and verifying, model explaining and executing and tenant managing and accessing controlling are specially covered.

The invention discloses a calculation method and device for a contribution degree of a training data set, equipment and a storage medium, and relates to the field of financial science and technology,and the method comprises the steps: obtaining each training data set of a training machine learning model; calculating a SHAP target value of a SHAP interpretation method of each feature in the training data set; and calculating the contribution degree of the training data set according to the SHAP target value of each feature in the training data set. According to the invention, the contributiondegree of each training data set is obtained through corresponding calculation of the SHAP target value of each feature in the training data set; the importance degree of each training data set in theprocess of training the machine learning model is evaluated through the contribution degree of each training data set, so that the training data set for training the machine learning model can be selected more accurately, and the data prediction accuracy of the machine learning model obtained by training is improved.

The disclosure relates to a method and device for analyzing data based on machine learning model interpretations. The method for analyzing the data based on the machine learning model interpretationscomprises the following steps: obtaining and displaying model interpretation content, wherein the model interpretation content comprises at least one of a model structure interpretation, model featureimportance and a model prediction interpretation; receiving a data analysis request made by a user resulted from at least one of feature names displayed in the model interpretation content; separately calculating data distributions of all features of each of the at least one feature name in a training sample; and visually outputting the data distributions.

The invention discloses a self-paced-cooperative training learning method. The method comprises the following steps of: obtaining data, from two visual fields, of a target problem, and initializing a model, wherein the data comprises a small amount of labeled data and a large amount of unlabeled data; respectively determining corresponding optimization targets on the two visual fields; embedding a self-paced regular term in a loss function of each visual field so as to realize steady learning under the visual field; associating the two visual fields through a regular term; obtaining a multi-visual field semi-supervised self-paced-cooperative training model which is embedded into a steady learning mechanism and has model interpretation; and obtaining high-quality labeling of the unlabeled data by applying the small amount of labeled data and large amount of unlabeled data in a target field and a semi-supervised multi-visual point learning model, and obtaining reliable learning devices under the two visual fields at the same time. The invention aims at providing a steady learning model with a replacement mode for the traditional cooperative training algorithms to ensure that data lack of labeling in the target field can obtain more correct and high-quality labeling.

The invention provides a machine learning model interpretation method and device, and the method comprises the steps: determining the model classification of a machine learning model; Explaining features in the machine learning model by adopting a model explanation method matched with the model classification; And comparing the explanation information of the feature obtained by explanation with reference explanation information to obtain an explanation effect of the feature. According to the machine learning model interpretation method provided by the invention, by understanding the decision basis of the machine learning model, the characteristic of high value of the machine learning model is screened out, and the characteristic screening quality of the machine learning model is improved.

The invention discloses a marking and explaining system and method for a large-data analysis model. The marking and explaining system comprises a model marking unit, a model managing unit, a model explaining unit, a model compiling unit and a user unit. The model marking unit is used for marking model metadata; the model managing unit is used for examining and approving the model, metadata managing and indexing; the model explaining unit is used for analyzing user operation and DAG relationship diagram converting; the model compiling unit is used for building the model dependency relationship, downloading a dependent library and jointly compiling; the user unit is used for identifying user identification and design achieving, managing and submitting of a user to the model. By means of the marking and explaining system and method, marking and explaining of the model in the large-data analysis process are achieved, operation when the user creates operation is convenient, and the calling process and the explaining process of the model are quickened.

Embodiments of the invention provide a model interpretation method and a device based on cooperative game, and electronic equipment, the method is applied to the electronic equipment, and the electronic equipment comprises a trained machine learning model for prediction or classification. The method comprises calculating a contribution value of each target feature selected from a plurality of features in at least one input test sample to the output result according to the output result of the machine learning model, so as to explain the output result of the machine learning model according tothe contribution value of each selected target feature to the output result. According to the method, a machine learning model is combined with a cooperative game theory, and a prediction result of the machine learning model is explained by calculating contribution degree values of target features, including a nonlinear model and a linear model. Meanwhile, the method can achieve the interpretationof the machine learning model for a single target feature, and also can achieve the interpretation of the overall prediction result of the machine learning model.

The embodiment of the invention provides a service processing method and device, equipment and a storage medium. The method comprises: processing service data corresponding to a to-be-processed service request by adopting a service processing model to obtain the prediction result of the target variable of the service processing model; and interpreting the prediction result of the target variable by adopting the preset local interpretation mode to obtain the local service interpretation corresponding to the prediction result of the target variable. Visibly, in the embodiment of the invention, the machine learning model and the model interpretation are combined so that a machine learning model is made to be a white box, related service interpretation information about the machine learning model is obtained, the interpretability of the machine learning model is improved; on one hand, the credibility of the model is improved, and the utilization rate of the machine learning model in the technical field with the high requirement for the interpretability of the model is increased; and on the other hand, the user group suitable for the to-be-processed service can be determined in combination with the service interpretation information.

A system, method, and program product for interpreting, optimizing, and customizing data mining models through the use of statistical techniques that utilize diagnostic measures and statistical significance testing. A data processing system is disclosed that includes a data mining system for mining data from a data warehouse in accordance with a data model, wherein the data model defines how data groups can be partitioned; and a data group analysis system that calculates a set of diagnostic measures and performs statistical significance tests for a defined data group.

The present invention provides a framework for explainability of a machine learning-based decision-making system. The framework calculates the directional contribution and sensitivity of each feature for each prediction. In addition, the framework provides decision rules to explain each prediction made by the decision-making system. Furthermore, the framework displays a readable explanation of the decisions made by the decision-making system via mapping the model explanation to the business context.

The embodiment of the invention provides a method and device for evaluating a model interpretation tool, and the method comprises the steps: training a first model by adopting a plurality of first training samples so as to obtain the first model with a first parameter group; acquiring the first performance value of the first model with the first parameter group based on a plurality of test samples; based on the plurality of first training samples and the first parameter group, obtaining importance sequences of a plurality of features through a model interpretation tool; replacing the feature values of the features except the first n features of the importance sequence in each first training sample with the same predetermined value to obtain a plurality of second training samples; trainingthe first model by using the plurality of second training samples to obtain a first model with a second parameter group; acquiring the second performance value of the first model with the second parameter group based on the plurality of test samples; and calculating a difference value between the first performance value and the second performance value for evaluating the model interpretation tool.

The invention discloses a model interpretation method, apparatus and device, and a readable storage medium. The method comprises the steps of performing interpretation processing on a model predictionresult of a target user by adopting a preset interpretation algorithm to obtain an interpretation result, generating a plurality of interpretation text contents according to the interpretation result, and analyzing each interpretation text content by adopting an acceptability discrimination model to obtain an acceptability corresponding to each interpretation text content, the acceptability discrimination model being obtained by training feedback data of historical interpretation text contents based on historical users, and selecting a target interpretation text content from the interpretation text contents based on the acceptability and outputting the target interpretation text content. According to the method, the actual application effect of model interpretation is improved.

The invention provides an interpretation scheme for lift tree model transfer learning, which can interpret the transfer learning based on the lift tree model and can support local interpretation and global interpretation of the transfer model.

The invention belongs to the field of intelligent auditing and particularly relates to an intelligent auditing method based on machine learning model interpretation. As for most of existing auditing methods adopt a manual mode for auditing, laws, regulations and rules needing to be followed in the auditing process are updated quickly. In order to solve the problems that the auditing accuracy and reliability are influenced only by brain storage of auditors and the auditing risk is high in the prior art, the invention provides the following scheme that the method comprises the following steps: S1, inputting a to-be-audited file or data to the input end of an intelligent auditing system based on machine learning model interpretation; S2, displaying an auditing process by utilizing a machine learning model while auditing the input file or data; S3, performing visualization and business rule output on a machine learning result through a model interpretation function of machine learning. According to the invention, the auditing accuracy and reliability can be improved, and the life cycle management of the auditing rule can be perfected.

The invention discloses a model interpretation method and equipment and a readable storage medium. The model interpretation method comprises the following steps of obtaining each model input feature corresponding to a preset black box model, inputting the prediction data set corresponding to each model input feature into a first hash coding model optimized based on each preset sample category; carrying out hash coding on the prediction data set; obtaining a first hash coding result, inputting the prediction data set into a second hash coding model optimized based on each output sample categoryoutput by the preset black box model; and performing hash encoding on the prediction data set to obtain a second hash encoding result, and further determining feature confidence corresponding to eachmodel input feature based on each bit difference degree between the first hash encoding result and the second hash encoding result. The technical problem that the model interpretation effect is pooris solved.

The invention discloses a DOA estimation method based on a machine learning algorithm XGBoost. The DOA estimation method comprises the following steps: acquiring a noisy array signal to obtain a covariance matrix; obtaining a data set according to the covariance matrix; constructing a training set and a test set according to the data set; training the training set by adopting an XGBoost algorithmmodel, and calculating optimal parameters of the model; and predicting the test set according to the optimal parameters of the XGBoost algorithm model. According to the DOA estimation method based onthe machine learning algorithm XGBoost provided by the invention, the prediction speed and precision are improved; meanwhile, the method has the advantages that the method is not easily influenced byabnormal values, a large amount of training data is not needed and the model interpretability is good.

The invention discloses a system and method for generating SPARQL query statements in the medical field, and belongs to the field of machine translation. The system comprises a generator which takes aquery template library and a knowledge base as input and is used for extracting entities and attributes from the knowledge base and filling the entities and attributes into a Chinese question template and an SPARQL query template to generate a training set; the word segmentation module is used for carrying out word segmentation processing on the Chinese questions in the training set and forwarding a word segmentation result to the learner; performing word segmentation processing on the target Chinese question, and forwarding a word segmentation result to an interpreter; the learner is used for training the neural network model according to the Chinese training set after word segmentation to obtain a trained model; and the interpreter is used for predicting the target Chinese questions after word segmentation by using the trained neural network model to obtain predicted SPARQL query statements, so that complex statistical and manual models are not used any more, and the medical healthquery Chinese questions are directly converted into the SPARQL query statements.

Explanation apparatus and methods are described. In one aspect, an explanation apparatus includes processing circuity configured to access a source instance which has been classified by a machine learning model; create associations of the source instance with a plurality of training instances; and process the associations of the source instance and the training instances to identify a first subset of the training instances which have less relevance to the classification decision of the source instance by the machine learning model compared with a second subset of the training instances; and an interface configured to communicate information to a user, and wherein the processing circuitry is configured to control the user interface to communicate the second subset of the training instances to the user as evidence to explain the classification of the source instance by the machine learning model.

The invention discloses an object-oriented multi-level interpretation method and system for a live-action three-dimensional model. A to-be-interpreted live-action three-dimensional model is divided into a terrain surface part and a three-dimensional ground object part; the terrain surface part is classified by adopting an object-oriented image classification method based on the orthoimage of the terrain surface part; independent ground object monomerization processing is carried out on the three-dimensional ground object part, and based on the characteristics of independent ground object monomers, a machine learning algorithm is adopted to classify the three-dimensional ground object part; and classification results of the terrain surface part and the three-dimensional ground object part are integrated to obtain an interpretation result of the whole live-action three-dimensional model. According to the method, the live-action three-dimensional model is divided into two layers to be classified respectively, so that the model interpretation difficulty is reduced; an object-oriented model interpretation method is adopted, so that other salt and pepper noise phenomena generated in a classification process based on a patch algorithm can be improved; based on a multi-dimensional feature fusion strategy of geometry, texture and spectrum, the accuracy and robustness of a classificationresult are improved.

The invention discloses a hybrid indexing method based on big-data model metadata. The method comprises the following steps: S1, extracting hybrid index metadata, wherein the metadata of a hybrid index is extracted according to the big-data model metadata and model interpretation and operation characteristics, and a value of 1 or 0 is assigned to the metadata according to an attribute value of the metadata; S2, constructing or updating the hybrid index, wherein a global hash function is utilized to construct or update the hybrid index; S3, storing the hybrid index, wherein all parts of the hybrid index are stored into a memory, a cache and a disk according to hybrid index features, and index contents are sequentially and concurrently retrieved according to a query request; and S4, retrieving the hybrid index, wherein a retrieval algorithm is constructed according to big-data model features and hybrid index characteristics, and the different parts of the hybrid index are retrieved at the same time. According to the method, the big-data model metadata and the model characteristics are tightly integrated, a high-efficient and accurate model indexing technology is provided, the retrieval speed is increased, and the convenience of big-data model using is improved.

The invention discloses a health state detection method and device based on model interpretation, and a readable storage medium. The health state detection method based on model interpretation comprises the following steps: acquiring to-be-detected user data corresponding to a target user, inputting the to-be-detected user data into a preset health state prediction model, and detecting the healthstate of the target user to acquire an initial detection result of a health state; and performing model prediction interpretation on the initial detection result of the health state to obtain a healthstate influence factor corresponding to the initial detection result of the health state, and generating a target health state detection result corresponding to the target user based on the health state influence factor and the initial detection result of the health state. According to the invention, the technical problem of low confidence degree of health state detection is solved.

The invention provides a stand growth and yield model dynamic state management method based on an encryption technique, and relates to the technical field of forestry resource management and forestry informatization. The method comprises the following steps: S1, a stand growth and yield model table is built; S2, the symmetric encryption technique DES (data encryption standard) is used for performing encryption to a model stored in the stand growth and yield model table, and a cipher text generated through encryption is written in the stand growth and yield model table; S3, the symmetric encryption technique DES is used for performing decryption to the cipher text read in the stand growth and yield model table, and the plaintext generated through decryption is displayed to a user; S4, the structured query language (SQL) is used for performing analysis to a model in the plaintext. The method builds a stand growth and yield model table, closely related to the origin, of the same area, uses the encryption technique to perform management to the access of data in the model table, provides a fast and simple model interpretation method, and finally realizes flexile and safe dynamic management of the stand growth and yield model.

The invention discloses a model feature screening method and device and a readable storage medium. The model feature screening method comprises the following steps: step 1, establishing a corresponding ith model by adopting an ith group of candidate features, wherein i is greater than or equal to 1; step 2, carrying out model interpretation on the ith model to obtain a feature contribution corresponding to each feature in the ith group of candidate features; step 3, screening the i-th group of candidate features according to the feature contribution corresponding to each feature, and excludingfeatures which do not meet a preset condition to obtain an i + 1-th group of candidate features; and step 4, taking the (i + 1) th group of candidate features as the ith group of candidate features,and repeatedly executing the step 1 to the step 3 until all the features which do not meet the preset condition are excluded to obtain target features. By adopting the scheme, effective model featurescan be accurately screened out.

The invention discloses a model interpretation method and device and a readable storage medium. The method comprises the steps: obtaining to-be-interpreted sample data, and distributing the sample data to all execution units in a cluster; broadcasting a to-be-interpreted model to each execution unit, so that each execution unit calculates the contribution degree of the target sample data to the prediction result of the to-be-interpreted model based on a preset model interpretation algorithm after receiving the to-be-interpreted target sample data. According to the method, the model interpretation efficiency is improved, and the method can be used for processing model interpretation tasks of more complex service scenes.

The embodiment of the invention provides a method and device for evaluating a model interpretation tool, and the method comprises the steps: training a first model through employing a plurality of training samples, so as to obtain the first model with a first parameter group, and the first model is a self-interpretation model; based on the self-interpretation of the first model with the first parameter group, obtaining a first importance sequence of the plurality of features; based on the plurality of training samples and the first parameter group, obtaining a second importance sequence of theplurality of features through a model interpretation tool; and determining the similarity between the second importance sequence and the first importance sequence so as to evaluate the model interpretation tool.