Abstract: To automate time series forecasting machine learning pipeline generation, a data allocation size of time series data may be determined based on one or more characteristics of a time series data set. The time series data may be allocated for use by candidate machine learning pipelines based on the data allocation size. Features for the time series data may be determined and cached by the candidate machine learning pipelines. Predictions of each of the candidate machine learning pipelines using at least the one or more features may be evaluated. A ranked list of machine learning pipelines may be automatically generated from the candidate machine learning pipelines for time series forecasting based upon evaluating predictions of each of the one or more candidate machine learning pipelines.

Abstract: An example operation may include one or more of storing a hierarchical time-series data set in memory, initially training a first time-series forecasting model based on a lower level of time-series data in the hierarchical data set, training a second time-series teaching forecasting model based on an upper level of time-series data from the hierarchical data set which includes an additional level of aggregation with respect to the lower level of time-series data, optimizing one or more parameters of the initially trained first time-series forecasting model based on predicted outputs from the trained second time-series forecasting model in comparison to predicted outputs from the initially trained first time-series forecasting model, and storing the modified first time-series forecasting model in the memory.

Abstract: A computer-implemented method, a computer program product, and a computer system for adaptive retraining of an artificial intelligence model. A computer system computes drift magnitude scores for respective drift functions. A computer system computes an aggregated data drift score for a data drift, an aggregated concept drift score for a concept drift, and an aggregated model drift score for a model drift. A computer system computes an overall drift score, based on the aggregated data drift score, the aggregated concept drift score, the aggregated model drift score, a predetermined data drift threshold, a predetermined concept drift threshold, and a predetermined model drift threshold. A computer system determines whether retraining of the artificial intelligence model is required, based on the overall drift score. A computer system performs the retraining of the artificial intelligence model, in response to determining the retraining of the artificial intelligence model is required.

Abstract: Selecting a time-series forecasting pipeline by receiving target variable time-series data and exogenous variable time-series data, generating a regular forecasting pipeline comprising a model according to the target variable time-series data, generating an exogenous forecasting pipeline comprising a model according to the target variable time-series data and the exogenous variable time-series data, evaluating the regular forecasting pipeline and the exogenous forecasting pipeline, selecting a pipeline according to the evaluation, and providing the selected pipeline.

Abstract: Anomaly detection in industrial dynamic process can include receiving a set of multivariate time series data representative of sensor data obtained over time. The set of multivariate time series data can be transformed into a set of signature vectors in an embedding space. A neural network can be trained to estimate a probability distribution of the set of signature vectors in the embedding space. Streaming data can be received. The streaming data can be appended with a previously stored time series data. The appended streaming data can be transformed into an embedding. The embedding can be input into the trained neural network, the trained neural network outputting a first probability distribution score. A second probability distribution score associated with the embedding can be determined based on a given proposed probability distribution. Anomaly score can be determined based on the first probability distribution score and the second probability distribution score.

Abstract: A system and a method of managing a manufacturing process includes receiving production data relating to the manufacturing process and determining an operational mode associated with the manufacturing process using historical, multivariate senor data. The method may further determine a recommended action to affect production based on the determined operational mode. The operational mode may be based on at least one of: a level of operation in a continuous flow process relating to a joint set of process variables, a representation of a joint dynamic of the set of process variables over a predefined length, and a joint configuration of an uptime/downtime of a plurality of units comprising a process flow.

Abstract: A method, computer system, and a computer program product for generating a custom corpus is provided. The present invention may include generating a domain graph. The present invention may also include gathering seed data based on the generated domain graph. The present invention may then include identifying domain related data based on the gathered seed data. The present invention may further include querying the domain related data. The present invention may also include creating word embeddings for the domain related data. The present invention may then include evaluating the domain related data.

Abstract: A computer implemented method of administering a complex system includes receiving multivariate data from a plurality of sensors of the system in an ambient state. Event sequences in the received multivariate data are identified. The multivariate event sequences are projected to a lower stochastic latent embedding. A temporal structure of the sequences is learned in a lower latent space. A probabilistic prediction in the lower latent space is provided. The probabilistic prediction in the lower stochastic latent space is decoded to an event prediction in the ambient state.

Abstract: A computer-implemented method for providing interpretable predictions from a machine learning model includes receiving a data structure that represents a hierarchical structure of a set of features (X) used by one or more predictive models to generate a set of predictions (Y). An interpretability model is built corresponding to the predictive models, by assigning an interpretability to each prediction Yi based on the hierarchical structure. Assigning the interpretability includes decomposing X into a plurality of partitions Xj using the hierarchical structure, wherein X=U1NXj, N being the number of partitions. Further, each partition is decomposed into a plurality of sub-partitions using the hierarchical structure until atomic sub-partitions are obtained. A score is computed for each partition as a function of the predicted scores of the sub-partitions, wherein the predicted scores represent interactions between the sub-partitions. Further, an interpretation of a prediction is outputted.

Abstract: To automate time series forecasting machine learning pipeline generation, a data allocation size of time series data may be determined based on one or more characteristics of a time series data set. The time series data may be allocated for use by candidate machine learning pipelines based on the data allocation size. Features for the time series data may be determined and cached by the candidate machine learning pipelines. Predictions of each of the candidate machine learning pipelines using at least the one or more features may be evaluated. A ranked list of machine learning pipelines may be automatically generated from the candidate machine learning pipelines for time series forecasting based upon evaluating predictions of each of the one or more candidate machine learning pipelines.

Abstract: A hybrid sensor can be generated by training a machine learning model, such as a neural network, based on a training data set. The training data set can include a first time series of upstream sensor data having forward dependence to a target variable, a second time series of downstream sensor data having backward dependence to the target variable and a time series of measured target variable data associated with the target variable. The target variable has measuring frequency which is lower than the measuring frequencies associated with the upstream sensor data and the downstream sensor data. The hybrid sensor can estimate a value of the target variable at a given time, for example, during which no actual measured target variable value is available.

Abstract: Embodiments of the invention are directed to collecting, by a computer system, sensor data of a manufacturing system, the sensor data being measured at intervals smaller than a time interval of a target measurement of the manufacturing system. The sensor data is determined to have a relationship to the target measurement. A synthetic target measurement is generated at an interval smaller than the time interval based on the relationship. An advance warning is automatically generated for the target measurement based on the synthetic target measurement within the interval smaller than the time interval.

Abstract: In various embodiments, a computing device, a non-transitory storage medium, and a computer implemented method of improving a computational efficiency of a computing platform in processing a time series data includes receiving the time series data and grouping it into a hierarchy of partitions of related time series. The hierarchy has different partition levels. A computation capability of a computing platform is determined. A partition level, from the different partition levels, is selected based on the determined computation capability. One or more modeling tasks are defined, each modeling task including a group of time series of the plurality of time series, based on the selected partition level. One or more modeling tasks are executed in parallel on the computing platform by, for each modeling task, training a model using all the time series in the group of time series of the corresponding modeling task.

Abstract: A computer implemented method includes generating a pipeline graph having a plurality of layers, each of the plurality of layers having one or more machine learning components for performing a predictive modeling task. A plurality of pipelines are operated through the pipeline graph on a training dataset to determine a respective plurality of results. Each of the plurality of pipelines are distinct paths through selected ones of the one or more machine learning components at each of the plurality of layers. The plurality of results are compared to known results based on a user-defined metric to output one or more leader pipelines.

Abstract: Machine logic to change steps included in and/or parameters/parameter value used in artificial intelligence (“AI”) pipelines. For example, the machine logic may control what types of data (for example, sensor data) are received by the AI pipeline and/or have the data is culled in the pipeline prior to application of a machine learning and/or artificial intelligence algorithm.

Abstract: A computer-implemented method for providing interpretable predictions from a machine learning model includes receiving a data structure that represents a hierarchical structure of a set of features (X) used by one or more predictive models to generate a set of predictions (Y). An interpretability model is built corresponding to the predictive models, by assigning an interpretability to each prediction Yi based on the hierarchical structure. Assigning the interpretability includes decomposing X into a plurality of partitions Xj using the hierarchical structure, wherein X=U1NXj, N being the number of partitions. Further, each partition is decomposed into a plurality of sub-partitions using the hierarchical structure until atomic sub-partitions are obtained. A score is computed for each partition as a function of the predicted scores of the sub-partitions, wherein the predicted scores represent interactions between the sub-partitions. Further, an interpretation of a prediction is outputted.

Abstract: A hybrid sensor can be generated by training a machine learning model, such as a neural network, based on a training data set. The training data set can include a first time series of upstream sensor data having forward dependence to a target variable, a second time series of downstream sensor data having backward dependence to the target variable and a time series of measured target variable data associated with the target variable. The target variable has measuring frequency which is lower than the measuring frequencies associated with the upstream sensor data and the downstream sensor data. The hybrid sensor can estimate a value of the target variable at a given time, for example, during which no actual measured target variable value is available.

Abstract: A system and a method of managing a manufacturing process includes receiving production data relating to the manufacturing process and determining an operational mode associated with the manufacturing process using historical, multivariate senor data. The method may further determine a recommended action to affect production based on the determined operational mode. The operational mode may be based on at least one of: a level of operation in a continuous flow process relating to a joint set of process variables, a representation of a joint dynamic of the set of process variables over a predefined length, and a joint configuration of an uptime/downtime of a plurality of units comprising a process flow.

Abstract: A user trajectory graph may be constructed based on spatio-temporal data. A mobility pattern may be extracted from the user trajectory graph. Users may be clustered into groups, wherein the users in a same group possess similar feature values in the mobility pattern, and the users in different groups have different feature values, to identify personas and location sets. A distribution model may be constructed that models user timing and location preference, wherein an outcome indicates a preference for a particular time bin on a particular day for a particular location.

Abstract: A method, computer system, and a computer program product for generating a custom corpus is provided. The present invention may include generating a domain graph. The present invention may also include gathering seed data based on the generated domain graph. The present invention may then include identifying domain related data based on the gathered seed data. The present invention may further include querying the domain related data. The present invention may also include creating word embeddings for the domain related data. The present invention may then include evaluating the domain related data.