Analyzing sequence data using neural networks

A computer-implemented system, comprising:

• a computer processor;
• and
• a computer readable non-transitory storage medium storing instructions thereon, the instructions when executed by a processor cause the processor to perform the steps of:
• receiving a plurality of sequences of data, wherein each sequence of data from the plurality of sequences represents a time series having a distinct length;
• for each of the plurality of sequences of data:
• providing the sequence of data as input to a recurrent neural network, the recurrent neural network comprising one or more hidden layers of nodes;
• and
• extracting a feature vector representing output of a hidden layer of the recurrent neural network, wherein the feature vector is generated by the hidden layer responsive to providing the sequence data as input to the recurrent neural network;
• clustering the extracted feature vectors to determine a plurality of clusters of feature vectors;
• for each cluster, determining one or more salient features characterizing the cluster;
• and
• sending information describing the clusters to a client device for presentation via a user interface.

A computer-implemented method comprising:

• receiving a plurality of sequences of data, wherein each sequence of data corresponds to sensor data generated by a sensor, and wherein each sequence of data represents a time series having a distinct length;
• determining a plurality of groups of the sequences of data;
• for each group of sequences of data, determining one or more salient features characterizing the group of sequences of data, the determining comprising:
• receiving a set of explanative features;
• for each explanative feature from the set of explanative features, determining a feature score based on a comparison of a measure of similarity of the explanative feature within the group of sequences of data with a measure of similarity of the explanative feature in a second group of sequences of data;
• and
• identifying one or more explanative features from the set of features as salient features based on the feature scores;
• and
• receiving information identifying a particular sensor;
• determining a particular group of sequence of data corresponding to the particular sensor;
• and
• predicting an event associated with the particular sensor based on the salient features characterizing the particular group of sequences of data.