TELECOM RAN
Optimization​

  • Analyze 4G and 5G call trace data to compute optimal base station radio parameters and detect anomalies​ (~30,000+ radio signaling events/second)​

  • Identify problematic areas and suggest optimal radio parameters​

  • Instant collection of radio data, enrichment, correlate and analyze huge amounts of records into actionable data structures, extracting and storing information in the RAN data lake​

  • Near-realtime RAN data processing for predicting events and instant optimization​

  • Vendor agnostic, can be used for traditional vendors like Huawei, Ericsson or Open RAN like Mavenir​

TELECOM RAN
Over the Air Drive Testing (OTA-DT)​​

  • Can complete/replace the traditional drive testing, reducing the number of trips up to 3x​​

  • Improves Customers QoE (with Revenue impact)​

  • Full API/SDK support for application developers and automations​

  • Complements existing solutions with unstructured data capabilities​​

  • Fast integration in customer environments to connect existing data and services​

  • SaaS Private/Hybrid Cloud​

  • Can perform Network audits for seamless performance across MNOs of the same GROUP​

TELECOM RAN
OUR PROPOSITION​

Step 1 - deploy BigCONNECT to:​

Collect real-time location data from MDT messages ​
Work with location data, no regression or triangulation​

Deep-learning AI model to predict geo-location based on CELL_ID, TA and RSRP​

Predict location when MDT location is not available​

Step 2 - fine-tune BigCONNECT’s AI models:​

Connect the system to live data​

Create accurate UE geo-location model​

Determine optimal antenna params based on aggregated UE locations, TA, measurement reports and historical traffic patterns​

Report anomalies​

Deliver predefined or custom reports​

TELECOM RAN
EXAMPLE

Each red dot is one UE​

Each UE placed on the map is based on geolocation (lat/long and location predicted with AI)​

Triangulation is avoided because of low resolution​

20x20m geo raster is overlayed on the map​

Each road to be evaluated by “instant drive testing” is defined by yellow raster (from CDG-BV. AVIATORILOR)​

The intersection between UEs and yellow raster coordinates will create the sum of measurements for the road alone ​

The UEs ‘on the road’ are identified as the ones in greed contour​

Only the green ones will be analyzed for the road drive testing.​

Everything else is useful for indoor and performance analytics, like traffic heat maps or outlier identification (anomalies detection)​

TELECOM RAN
BENEFITS

Instant collection of radio data with automatic enrichment & correlation​Turn radio data into actionable structures​

Create a RAN data lake​

Manage the complete data lifecycle, before and after optimization from collection to AI/ML to visualization​

Completes traditional drive testing and reduces the number of trips up to 3x​

The AI assisted network solution Improve Customers QoE (with Revenue impact), create Efficient Network density (with Capex impact), Reduce power consumption (Opex impact) with high impact for migration to 5G​

Complements existing solutions with unstructured data capabilities​

Fast integration in customer environments to connect existing data and services​

Offers Network audits, for seamless performance across MNOs of the same GROUP​

Vendor agnostic, can be used for traditional vendors, like Huawei, Ericsson or Open RAN like Mavenir​

Driverless drive testing
Optimization​

  • Analyze 4G and 5G call trace data to compute optimal base station radio parameters and detect anomalies​ (~30,000+ radio signaling events/second)​

  • Identify problematic areas and suggest optimal radio parameters​

  • Instant collection of radio data, enrichment, correlate and analyze huge amounts of records into actionable data structures, extracting and storing information in the RAN data lake​

  • Near-realtime RAN data processing for predicting events and instant optimization​

  • Vendor agnostic, can be used for traditional vendors like Huawei, Ericsson or Open RAN like Mavenir​

Driverless drive testing
Over the Air Drive Testing (OTA-DT)​​

  • Can complete/replace the traditional drive testing, reducing the number of trips up to 3x​​

  • Improves Customers QoE (with Revenue impact)​

  • Full API/SDK support for application developers and automations​

  • Complements existing solutions with unstructured data capabilities​​

  • Fast integration in customer environments to connect existing data and services​

  • SaaS Private/Hybrid Cloud​

  • Can perform Network audits for seamless performance across MNOs of the same GROUP​

osint
PROBLEM STATEMENT​

Intelligence Agencies across the world face significant challenges to identify, gather and interpret content from various sources.​​

Extracting and summarizing meaningful and relevant information from huge amounts of data is a big challenge.​​

Social Networks are tightening their data-collection policies, restricting more and more automated data collection capabilities​​

Predicting outcomes from existing data points is often not possible or too difficult​​

Deriving intelligence from existing data sources is time-consuming and performed manually​

osint
OBJECTIVES

To collect data from all available sources, both online and in-house​​

To identify information of interest and how it’s connected (persons/groups, associates, organizations, locations, phone numbers etc.)​​

To find what you need in a matter of seconds​​

To predict what will happen next so you can prevent​​

To get the exact big-picture of the situation​

osint
THE LIFECYCLE​

1. Planning​

Identify data sources​
Determine feasible access and acquisition methods​

2. Acquisition​

Data acquisition​
Language translation if necessary​
Data extraction​ Early detection (real-time processing)​

3. Enrichment​

Link Analysis, Spatial Analysis​
Text, Images, Videos, Audio​
Documents, Web Content, Social Content, Dark Web Content​
Object detection, speech recognition, entity recognition, relationship extraction, sentiment analysis​
Image/video/audio analytics​

4. Analysis​

Knowledge discovery: data mining, search, data visualization, data science​Prediction​
Collaboration​
Watch list​
Continuous monitoring​

5. Dissemination​

Intelligence reports​
Real-time alerts​
Automated actions​
Dashboards​
Enforcement
​Gap analysis​

osint  
TECHNOLOGY

Highly secure, multi-model, massively-scalable datastore​​

OSINT and SOCINT data collectors and rotating proxy systems​​

Store massive amounts of data, never delete a single thing​​

Sub-second interrogation capability​​

Trainable AI models​

​Integrates with your existing systems and databases​

Low-code ecosystem, extensible, no vendor lock-in !​

osint
BENEFITS

Efficient and effective information retrieval from online and in-house data sources​​

Information extraction from all types of data: video, image, audio and multilingual textual content​​

Real-time event correlation, detection and visualization​​

Strategize on possible future trends and hot-topics before they happen​​

Central knowledge system that grows over time​

Call center
INTELLIGENT CONTACT CENTER

Customer satisfaction (CSAT) is the ultimate measure of customer loyalty

Self-service is the biggest opportunity for organizations to reduce costs

Customers abandon self-service when they lose confidence in their ability to resolve an issue without the help of another person.

Low-effort customer interaction is the main driver for growth and acquisition:

A low-effort, faster resolution process has three critical functions:
- Focusing on customer outcomes
- Managing customers’ perception of the outcome
- Staying ahead of foreseeable issues

Call center
BIGCONNECT PLATFORM

Deploy AI to recognize a customer’s intent, to provide answers, understanding requests that aren’t a part of a predetermined menu of options

Quickly identify the customer, their products, services and previous interactions

Accurately respond with relevant answers

Automate tasks like paying a bill, completing orders, and providing instructions to customers without channel switching

Send automatic alerts when an issue is solved

Proactively alert customers when something is wrong

Call center
BENEFITS

Provide Clarity, Confirmation, Credibility for your customers

Significantly reduce contact center load during major incidents

Achieve low-effort customer interaction that reduce costs and boosts sales

Become proactive instead of being just reactive

Guide customers to the right service channel

Super-fast resolution times

ran Engineering 
4G/5G RAN​

Instant collection of radio data, enrichment, correlation and analysis of huge amounts of records into actionable data structures, extracting and storing information in the RAN data lake;

Manages the complete data lifecycle, before and after optimization from collection to AI/ML to visualization;

Self-service & no-code (zero touch) for technical and business users;

Processing near real time RAN data for predicting events and fast/instant optimization;

Can complete/replace the traditional drive testing and reduce three times the number of trips;

The AI assisted network solution improves Customers QoE (with Revenue impact), creates Efficient Network density (with Capex impact), reduces Power consumption (Opex impact) with high impact for migration to 5G.

ran Engineering 
4G/5G RAN​

Full API/SDK support for application developers and automations;

Complements existing solutions with unstructured data capabilities;

Fast integration in customer environments to connect existing data and services;

SaaS Private/Hybrid Cloud;

Provides Network audits, for seamless performance across MNOs of the same GROUP;

The solution is vendor agnostic, can be used for traditional vendors, like Huawei, Ericsson or Open RAN like Mavenir. The unstructured data will be mapped for each parameters, otherwise lost after each call;

The BigConnect implementation outcomes will impact Coverage, Mobility, Capacity and Interference;

ran Engineering 
OUR PROPOSITION​

CSP CURRENT

Always build regression models to predict TA from RSRP
Triangulate
Sensitive to geography  
Location might not be accurate or incomplete (no road for drive testing)
May not be reliable for further optimizations

BigConnect

Collect location data from MDT messages directly
Work with location data, no regression or triangulation
Train a deep-learning model to predict geo-location based on CELL_ID, TA and RSRP
Use the model to predict location when MDT location is not available

ACTIONS

Connect the system to live data • Create accurate UE geo-location model
Determine optimal antenna params based on aggregated UE locations, TA, measurement reports and historical traffic patterns
Report anomalies
Deliver predefined or custom reports

FRAUD PREVENTION
TYPES OF FRAUD

Money Laundering – Moving from People to Corporations, Shell Companies and Back
Credit Card – Present or Not Present/Skimming
Identity Theft – Synthetic IDs and Account Takeovers
Combinations – Identity theft + Credit Card + Money Laundering Fraud we don’t know about yet...

FRAUD PREVENTION
COMMONALITY

Smurfing: moving money between small accounts
Transactions: Amounts, Locations, Types of Goods, Stores
Actors: People, Companies, Goods and Services
Locations: Physical, Corporate identities, Zip Codes
Devices: Mac Address, IP, IMEI, IMSI etc

FRAUD PREVENTION
STAGES

GUARD

Risk Pattern Monitoring

Identify data sources​
Semi / Fully automated
Looks for risk patterns “we have seen before”
Write programmatic rules
Fraudsters constantly working to “crack the code”

DISCOVER

Investigation

Expert + interactive
Link analysis
Case analysis
Charts / Dashboards
Connect with other data sources

ANTICIPATE

AI + ML

Learn from existing patterns
Look at new data sources (social media, web)
Semi / Fully automated
Looks for risk patterns “we have not seen before”
Looks for recurring patterns
Does not need rules
Difficult to “crack”

FRAUD PREVENTION  
GRAPHS IN FINANCIAL SERVICES

Traditional fraud analytics looks for outliers but cannot detect patterns (e.g. Fraud Rings)

Pattern matching
Relationship & association analysis
Real-time monitoring and decisions
Can adapt to any change or data

FRAUD PREVENTION
WHO ARE TODAY'S FRAUDSTERS

Organized in groups
Synthetic identities
Stolen identities
Hijackd devices

FRAUD PREVENTION
GRAPHS IN FINANCIAL SERVICES

DEPENDENCIES

Order of operations

CLUSTERING

Find things closely related to each-other (e.g. friends, affiliations, groups)

SIMILARITY

Find similar paths or patterns

MATCHING / CATEGORIZING

Highlight variant of dependencies

CENTRALITY

Find the most connected or relevant nodes

FLOW / COST

Find distribution problems and identify optimizations