System and method for automated insight curation and alerting

A method for automated insight curation and alerting implemented by a computer network comprising instructions that, when executed, enable the computer network to automatically generate insight curation and alerting, comprising

• receiving, via a signal receiver, an input signal in the form of time series from IoT devices;
• normalizing, via the signal receiver, the input signal through density-based kernels to generate a byte stream;
• passing, from the signal receiver, the byte stream to a message broker to store the byte stream in a distributed manner;
• passing, from the message broker, the byte stream to a decision engine comprising machine learning algorithms for forecasting;
• computing, via the decision engine, a normal range for behaviors of the byte stream;
• detecting, via the decision engine, anomalies in the byte stream;
• generating, via the decision engine, insight signals;
• and
• reporting, via a signal sender, the insight signals with supporting information;
• wherein the decision engine comprises an anomaly detection module to detect anomalies in the byte stream,
• wherein the anomaly detection module is configured to fetch the byte stream using windows, thereby generating a chunked signal,
• wherein the chunked signal is processed through a series of functional transformations via machine learning algorithms,
• wherein the functional transformations are implemented by a predictor, an anomaly detector, a root cause analyzer, an insight generator, a retainer and a modeling layer,
• wherein the predictor is configured to generate a predicted time series and a predicted driver based on historical patterns, wherein the predicted driver is a time series,
• wherein the anomaly detector is configured to find and flag the anomalies to generate a flagged anomaly signal, and
• wherein the root cause analyzer is configured to generate a numerical assessment by using the predicted driver, the input signal and the flagged anomaly signal.

A method for enabling businesses to spot, explore, and exploit anomalies and patterns in a dataset implemented by a computer network comprising instructions that, when executed, enable the computer network to automatically generate insight curation and alerting, comprising:

• receiving, via a signal receiver, an input signal in the form of time series from IoT devices;
• normalizing, via the signal receiver, the input signal through density-based kernels to generate a byte stream;
• passing, from the signal receiver, the byte stream to a message broker to store the byte stream in a distributed manner;
• passing, from the message broker, the byte stream to a decision engine comprising machine learning algorithms;
• computing, via the decision engine, a normal range for behaviors of the byte stream;
• detecting, via a hotspot identifier in the decision engine, business areas that are growing or declining by highlighting products or business units which have an impact by comparing the product or business unit against its peers as well as its past performance;
• detecting, via a breakout detector in the decision engine, a trend setting moments in businesses and generating an alert;
• detecting, via an outlier detector in the decision engine, anomalies in the dataset;
• detecting, via an early warning detector in the decision engine areas of the business that do not meet target outcomes;
• detecting, via a mix changes detector in the decision engine, changes in a business mix signalling an underlying change in business dynamics;
• generating, via a pare to alert generator in the decision engine, pare to alerts after detecting changes in a business mix;
• identifying, via a milestone identifier, milestones;
• detecting, via a trend reversal detector in the decision engine, changes in trends on critical business metrics;
• and
• detecting, via a non-occurrence detector in the decision engine, null value in the dataset;
• reporting, via a signal sender, insight signals with supporting information;
• wherein the anomalies comprise items, events and observations that do not conform to an expected pattern or other items;
• wherein the decision engine comprises an anomaly detection module to detect anomalies in the byte stream,
• wherein the anomaly detection module is configured to fetch the byte stream using windows, thereby generating a chunked signal,
• wherein the chunked signal is processed through a series of functional transformations via machine learning algorithms,
• wherein the functional transformations are implemented by a predictor, an anomaly detector, a root cause analyzer, an insight generator, a retainer and a modeling layer,
• wherein the predictor is configured to generate a predicted time series and a predicted driver based on historical patterns, wherein the predicted driver is a time series,
• wherein the anomaly detector is configured to find and flag the anomalies to generate a flagged anomaly signal, and
• wherein the root cause analyzer is configured to generate a numerical assessment by using the predicted driver, the input signal and the flagged anomaly signal.