ML&AI Applications

5G Analytics &Services

MNOs need to know – in real time – how users and applications are impacting the 5G network, or how adverse network conditions or unusual device behaviour is impacting users and applications. They also need to constantly monitor service experience metrics for each type of 5G service. Ampere AI 5G Analytics offer an open, multi-vendor, full-featured 5G Analytics that will interoperate with all 3GPP-compliant 5G Core components. Performing statistical or predictive analytics on collected data and applying AI/ML algorithms

Real-time network Analytics : Measure network load and detect congestion or other types of network performance anomalies.

Network slice instance load level computation and prediction

Load analytics information and prediction for a specific NF

Network performance computation and prediction

Congestion information – current and predicted for a specific location

Quality of service (QoS) sustainability reporting and predicting QoS change

Device Analytics: human users and machines will be connected to the 5G network. 5G operations require monitoring device connectivity, mobility and communications patterns. Abnormal device usage or unanticipated behavior by human users or smart machines could have a negative impact on network performance.

User equipment (UE) mobility analytics and expected behavior prediction
UE abnormal behavior / anomaly detection
UE communication analytics and pattern prediction

Service QoE: MNOs need to measure and track service experience to ensure that the 5G network is meeting user expectations and Service-Level Agreements (SLAs). NWDAF generated metrics will measure the current service experience by user, application type, device group or geographic location, and predict future changes.

Use cases:
Service experience computation and prediction for an application or UE group
Application service experience computation and prediction

Ampere AI Digital twins – the future of Supply Chain manufacturing

Ampere AI Supply Chain Manufacturing Digital twins uses AI, machine learning, and augmented reality that helps to gain information on how to improve manufacturing processes. It gives the ability to connect numerous machines, processes, and solutions into one working system. Digital twins generate highly valuable insights that help your organization to grow and meet industry challenges.
Challenge : The prediction of manufacturing lead time is a complex problem to solve which is influenced by constituents and constraints of the supply chain. Accurate and timely prediction of manufacturing lead time can assist in planning, streamlining operations, and increasing output and revenue. Poor assessment of manufacturing lead time can result in customer dissatisfaction, inventory mismatches, unutilized capacity and low resilience of the supply chain towards external factors.

Solutions : The Solution provides a machine learning (ML) based approach to accurately predict lead time at various stages for orders in the manufacturing supply chain industry. This solution provides an accurate assessment of manufacturing lead time along with the intermediate process dates for sub-processes/stages. The solution is validated on data from a payment card manufacturing supply chain. Below are key benefits of the solution.

Material Planning: To determine which materials we would need to create a BOM on ERP. Once the BOM is created, this module will tell whether there is any shortage expected in any of the material.

Marketing and Sales: AI powered price and Demand forecast

Order management: Agile production planning and scheduling

Capacity Management : Real time monitoring of OM KPIs and capture potential bottlenecks while executing an order.

End- to –End Transparency and faster decision making

New order automation: You can run a new order in the digital twin to know expected delivery data, intermediate process dates for sub-processes/stages, material shortage and capacity constrain to improve overall efficiency

Customer Care Operation Streamline and automation

Automate service operations for all touch points interactions

Convert touch points interaction to text,

Ampere AI model analyzes the text, understand and extract key contextual concepts of the conversation.

Match operational actions based on key concepts extracted

An action is triggered

Enable end to end automation for customer interactions

Predict ops flow and ticket resolution sequence

Machine learning based model enable right assignment to the right team at right sequence.

Enable automate ticket resolution, improve operational efficiency, reduce unnecessary operational time spent

Marketing AI

Use Case:

Competitive intelligence for marketing mix Analyze subscribers using known competitors to mobile services and apps

Personalized/targeted marketing campaigns with effectiveness/adoption measurement Target subscribers with interest/intention-based personalized messages and offers (vs. socio-demographic campaigns

Benefits:

Optimize marketing programs/budget: service awareness, retention and conversion/adoption
Leverage preferred communication channels
Maximize product/service adoption
Minimize service adoption cost
Avoid spamming subscribers
Increase Revenue, reduce cost and improve customer experience

Predictive Analytics Use Case

Incident Prediction, Maintenance Prioritization & Fault Description

Real-time Fault Prediction via Behavior Outlier Detection

Real-time Service Impact Prediction

Behavioral Outlier Detection

Advanced Anomaly Detection

Identify faults

Identify abnormalities

Predictive Maintenance: Fault Prediction

Predict equipment failure

Predict the equipment failure location – CPE versus network

Predictive Maintenance: Repair Prediction / Proactive Asset Maintenance

Predict repair events
Identify missed repairs

Behavioral Analytics

Real-time Personalized Targeting

eCommerce Attribution

Geometric (non-Collaborative Filtering) approach handles high-volume/velocity of data

Engagement Process

Business problem understanding

Use case and data definition

Success criteria definition

Data transfer complete

Data understanding Complete.

Data processing and data profiling

Exploratory data analysis

Feature Engineering

Model Building

Model tuning

ROI analysis

Results presentation

Align with stakeholders on business objectives

Document ‘as-is’ and expected ‘to-be’

Document business and technical use case

Detail data requirements

Define success criteria

Ensure technical, commercial, operational feasibility

Use case and success criteria discussion

Ensure availability of required data. Minimum 2 months of ‘good’ data needs to be available

Work with SMEs to understand data

Resolve all the data issues and ascertain feasibility of the use case

Develop data catalogue

Develop a project plan

SOW discussion and sign-off

Prepare data for analysis

Profile data, identify data trends and issues

Correlations and multi-collinearity analysis

Ensure sample stability

EDA results presentation

EDA feedback incorporation

Feature extraction

Identify key features

Work with SMEs to qualify features for modeling

Model building

Identify best-fit model

First-cut model discussions

Model tuning, incorporate model feedback, iterate working with SME

Finalize model

Build ROI or TCO models, operationalization model (PMO- FMO)

Final presentation Integration in customer environment