# US 12,015,531 B2 · Method for generating a quality of experience (QoE) index by way of ensemble of expectation scores

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 12,015,531 B2
- Document ID: US12015531B2
- Status: Granted
- Kind: B2
- Application number: 17/541,432
- Filed: 2021-12-03
- Published: 2023-06-08
- Granted: 2024-06-18
- Earliest claimed: 2021-12-03
- Assignees: Guavus, Inc.
- Applicants: GUAVUS, Inc.
- Inventors: Kuldeep Jiwani
- Web page: https://patentgrep.com/patent/US12015531B2

## CPC

- H04L 41/5067: Customer-centric QoS measurements
- H04L 41/142: using statistical or mathematical methods
- H04L 41/5009: Determining service level performance parameters or violations of service level contracts, e.g. violations of agreed response time or mean time between failures [MTBF]
- H04W 24/10: Scheduling measurement reports {; Arrangements for measurement reports}
- H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M)
- H04W: WIRELESS COMMUNICATION NETWORKS (broadcast communication H04H; communication systems using wireless links for non-selective communication, e.g. wireless extensions H04M1/72)

## Abstract

A method provides telecom operators a single Quality of Experience (QoE) index to collectively interpret service experience and network experience by way of ensemble of expectation scores. The method includes the steps of mapping Key Performance Indicators (KPIs) of time-series data into multiple probability spaces of statistical expectation functions thereby producing time-series expectation scores; applying vector geometry to said time-series expectation scores for each of said statistical expectation functions thereby producing an N-Dimensional probability vector; normalizing correlations of said N-Dimensional probability vector across said KPIs thereby producing a normalized N-Dimensional probability vector; and generating a N-Dimensional probability distribution from the normalized N-Dimensional probability vector to produce an ensemble function. The ensemble function is then be applied to time-series data thereby producing an ensemble index, which represents the QoE (Quality of Experience).

## Claims

### Claim 1

A method for providing telecom operators a single Quality of Experience (QoE) index to collectively interpret service experience and network experience by way of ensemble of expectation scores, the method comprising the steps of:
mapping Key Performance Indicators (KPIs) of time-series data into multiple probability spaces of statistical expectation functions thereby producing time-series expectation scores;
applying vector geometry to said time-series expectation scores for each of said statistical expectation functions thereby producing an N-Dimensional probability vector;
normalizing correlations of said N-Dimensional probability vector across said KPIs thereby producing a normalized N-Dimensional probability vector;
generating a N-Dimensional probability distribution from the normalized N-Dimensional probability vector to produce an ensemble function; and
applying the ensemble function to other time-series data thereby producing an ensemble index.

### Claim 2 (depends on 1)

The method of claim 1, wherein a time-series data is particular to a KPI, wherein a KPI is particular to a domain, said domain being at least one among service, customer segment, subscriber and flow.

### Claim 3 (depends on 2)

The method of claim 2, where said time-series data comprises any one or more of tonnage data, flow data, subscriber data, data throughput data, concurrency data, periodicity data, variance data duration data, or diversity data.

### Claim 4 (depends on 3)

The method of claim 3, where said step of mapping Key Performance Indicators (KPIs) comprises:
applying at least one said statistical expectation function to said time-series data thereby producing an expectation score time-series for each said KPI,
where said statistical expectation function comprise one among a threshold, a relative spike, a level shift, a volatility shift, a seasonal shift, a trend, a seasonal variance and a forecast.

### Claim 5 (depends on 4)

The method of claim 4, wherein said expectation score time-series is bounded from zero to a positive number, where a zero score represents a value that is closest to an expected value and a higher positive number represents a magnitude of deviation from the expected value.

### Claim 6 (depends on 4)

The method of claim 4, wherein said step of applying at least one said statistical expectation function produces a plurality of expectation score time-series for a set of multiple KPIs, wherein
said plurality of expectation score time-series is dimension E;
said set of multiple KPIs is dimension M; and
said expectation score time-series comprises scalar terms.

### Claim 7 (depends on 6)

The method of claim 6, where said step of applying vector geometry to said time-series expectation scores comprises:
calculating a mean and a variance for each said expectation score time-series for each said KPI of said set of multiple KPIs;
performing said calculating across said set of multiple KPIs to produce said N Dimensional probability vector from said mean of each said KPI, where N=M×E.

### Claim 8 (depends on 7)

The method of claim 7, wherein said step of normalizing correlations of said N-Dimensional probability vector comprises:
computing a covariance matrix of size (N×N) from said plurality of expectation score time-series,
wherein said normalizing correlations employs a Mahalanobis distance instead of a Euclidean distance to account for multi-variate data distributions of the N-Dimensional probability vector.

### Claim 9 (depends on 8)

The method of claim 8, wherein said variances occur along a diagonal of said covariance matrix, and cross-correlation occur outside said diagonal of said covariance matrix.

### Claim 10 (depends on 9)

The method of claim 9, wherein said step of generating a N-Dimensional probability distribution comprises incorporating Mahalanobis distance in a Gaussian kernel to produce said ensemble function.

### Claim 11 (depends on 10)

The method of claim 10, wherein said step of applying the ensemble function to said other time-series data comprises:
weighting the other time-series data by the ensemble function thereby generating (probability) p-values;
applying a negative log operation to the p-value to produce said ensemble index representing said QoE index; and
presenting the ensemble index.

### Claim 12 (depends on 11)

The method of claim 11, wherein the ensemble index represents an expectation deviation across all KPIs that incorporates specific statistical properties of said statistical expectation functions.

### Claim 13

A method for providing telecom operators a single Quality of Experience (QoE) index to collectively interpret service experience and network experience by way of ensemble of expectation scores,
where upon, collecting time-series data for multiple Key Performance Indicators (KPIs), the method comprising the steps of:
applying multiple statistical expectation functions to each said time-series data of a KPI from a set of the multiple KPIs thereby producing an expectation score time-series for each statistical expectation function of said KPI;
calculating a mean and a variance for each said expectation score time-series associated with said KPI for the set of multiple KPIs;
producing a N-Dimensional probability vector from said means representing the set of multiple KPIs,
where N=M(number of KPIs)×E(number of expectation functions);
computing a covariance matrix of size (N×N) from said plurality of expectation score time-series and said variances;
calculating a Mahalanobis distance using said N-Dimensional probability vector and said covariance matrix;
generating a N-Dimensional probability distribution that incorporates said Mahalanobis distance in a Gaussian kernel to produce said ensemble function.

### Claim 14 (depends on 13)

The method of claim 13,
where upon, collecting other time-series data for multiple Key Performance Indicators (KPIs),
the method further comprises the steps of:
weighting the other time-series data by the ensemble function thereby generating (probability) p-values;
applying a negative log operation to the p-value to produce said ensemble index representing said QoE index; and
presenting the ensemble index.

### Claim 15 (depends on 13)

The method of claim 13, wherein the weighting includes a matrix multiplication of the other time-series data by the covariance matrix.

### Claim 16 (depends on 13)

The method of claim 13, wherein time-series data is particular to a KPI, wherein a KPI is particular to a domain, said domain being at least one among Service, Customer Segment, Subscriber and Flow.

### Claim 17 (depends on 16)

The method of claim 16, where time-series data comprises any one or more of tonnage data, flow data, subscriber data, data throughput data, concurrency data, periodicity data, variance data duration data, or diversity data.