Advanced visual data mining and prediction system that enables you to analyse diverse, multi-dimensional and complex datasets.
Sign up for a demo

Better analysis of complex data most accurate representation

Deep Topological Analysis is an evolution of cutting edge data science. It combines both Topological Data Analysis and Deep Learning. This approach delivers provable performance gains over competing technologies.

Faster workflow

The simplicity of the DataRefiner platform saves time. Automated checks for data consistency, automatic selection of model parameters and other features make your data analysis pipeline much faster.

Explainable Artificial Intelligence

We have unique algorithms designed to make the process visual and accessible. Understand and communicate results with your organisation, clients and regulators.

How much do you know about your data?

Most data is not organised or structured making it hard to find the important patterns.

Your data is the biggest asset of your business

High level data segmentation can help any business thrive.

But what if its sources are user activities, sensors or texts?

Current business intelligence systems do not allow you to segment or discover patterns in such data. That’s where DataRefiner comes in: it provides tools to understand such data.

Transform your data into a data map

Imagine you could transform your data into a data map in just a few clicks and without specialist knowledge.

Deep Topological Analysis

Using machine learning, we group complex data into clusters, which allows users to discover hidden patterns.

State-of-the-art data segmentation

DataRefiner uses a unique algorithm to group complex data into clusters using deep topological analysis. It maps the segments and describes them to any business user without specialist training.

Visualise patterns in your data

You’re in control of the results: you can identify anomalies, predict behaviour, target audiences or eliminate engineering errors.

Customer and user activities

Extract personas from your user activities to understand their needs, experiences, behaviours and goals.

Internet of Things

Telemetry from your sensors provides important patterns for your equipment. Understanding of such patterns allow you to make improvements.

Bank transactions or crypto currency analysis

Spot anomalies or unusual patterns in your data. Perform deep analysis of such patterns using most advanced technology available on the market.

Discover things you haven't even started looking for

Patterns in your data are visualised to make them clear to anyone: from data scientists to business users.

Explainable AI

Achieve explainable results. DataRefiner visualises features in your data and makes them clear to data scientists or business users. It allows you to easily understand and communicate results within your organisation, to clients and regulators.

Сomply with GDPR

Make sure your machine learning systems are not biased against different groups of people. DataRefiner helps you verify the results of your algorithm.
“One of a kind platform that allowed us to understand our audience better. Advanced segmentation and sentiment analysis helped to present improved reports to our advertisers and drive our editorial content”
Mark Sear
Vice President, Audiences & Data, CNN (Turner Media)
“DataRefiner enabled us to provide deep, meaningful analysis of a large number of aircraft engines and sped up investigations of various issues, - particularly important because time was a critical factor.”
Faizan Patankar
Venturing and Incubation Lead, R² Data Labs, Rolls-Royce PLC

Interactive analytical platform for engineering and social sciences

DataRefiner user interface


Topological Data Analysis and Deep Learning

Deep Topological Analysis

A combination of Topological Data Analysis and Deep Learning gives the most accurate representation of your data.
Up to 50 million records

Large scale data analysis

Support up to 50 million records in one map. This allows analysis of very complex patterns in a large variety of cases.
Results in minutes

High performance

Using parallel computing, the platform can utilise up to 16 cores and with access to SSD it can process gigabytes of data in a minute.
Easy to use


The system takes care of meta-parameter optimisation automatically, the user just needs to upload the data and perform the analysis.
In a browser

Fully web-based platform

Perform the whole analysis in your browser, from data gathering to training, analysis and export of the result.

Interactive data analytics

WebGL allows us to use your GPU to interactively explore millions of data points in your browser in real time.
Automatic data preparation

Text and sensor data support

Text, sensor data and categorical parameters are automatically processed for better visualisation.
Cloud or on-premises

SaaS or on-premises

Use the flexibility of our managed cloud solution or deploy our platform on your network.

Access your project with API

Start your analysis using our API, track the training progress and export the results.


Easy to use

See your data like never before

DataRefiner represents your data as a set of clusters of similar content

Every dot is one or many data samples

Your data samples are presented as dots on a topological map. If two samples have similar values they can be put in a single point together.

Proximity of points indicates data similarity

Similar data samples are presented as points close to each other, different samples can be in an opposite part of the structure. The algorithm learns this structure and extracts clusters for analysis.

See the parameters of your data

Visualise parameter values of your data as colours on the map. It's easier to spot complex patterns like anomalies, malicious behaviour or unexpected types of user activity.

Simple presentation of complex data

Use 3D, 2D and 1D topological representation of your data

3D presentation

In many cases 3D data presentation helps to improve data pattern analysis and spot even finer details in your data.

Group data into clusters using a state-of-the-art approach

By using Topological Data Analysis and Deep Generative Networks we're able to extract the most complex patterns from the data and present them as separate clusters for further research by analysts.

Use traditional data analysis augmented with cluster analysis

Take the best from both worlds: complex segmentation and simple presentation

Plot your data on top of extracted clusters to find unexpected patterns

Explore assumptions about your data by overlaying extracted clusters with other parameters. Find and explore unexpected cases.

DataRefiner automatically detects the best parameters for visualisation

DataRefiner analyses your data and suggests the best graphs for exploration based on feature dependency, correlation and variability. Users can create any custom graph as well.

Explainable machine learning

Data is presented as clusters, each cluster is clearly explained.

Click on any cluster to understand it

Every cluster has a correlations panel, which presents the most representative parameters to describe it.

Click on any parameter to see its distribution across clusters

Clicking on the parameter name changes the colours of points on the map allowing you to visually understand data patterns.

Examine source data for any part of the structure

User is able to check the source data for any part of the map.

Understand and visualise text data

Explore the structure of your text and understand language associated with every product type: what people like and don’t like.

Topic extraction

Automatically extracted topics are presented as a cluster structure. Each cluster may be represented by more than one topic.

Key words for every cluster

See the keywords for every cluster, click any keyword to see its distribution across your data set.

Sentiment analysis included

Sentiment values are overlaid on the map for easier analysis, helping to spot clusters with highly positive and negative sentiment as well as to understand the data better.

Compare clusters and get precise descriptions for them

Dive deeper and explore clusters with sub-clusters, compare two clusters to understand why are they separate

Cluster criteria

DataRefiner is able to generate a precise rule on how to separate a selected cluster from the rest of the data, it's very useful for downstream tasks as well as deeper data understanding.

Cluster comparison

Compare two clusters and get a list of features which most distinguish them from each other.


Cutting-edge image clusterization using Topological Data Analysis
DataRefiner introduces a cutting-edge model that pushes the boundaries of unsupervised image clustering
April 3, 2024
GTA 6 Trailer Segmentation Using Deep Topological Analysis
We provide an illustration of video analytics utilizing Deep TDA (Topological Data Analysis)
December 18, 2023
Yield and Quality Revolution in Semi-Conductor industry: Deep Topological Analysis
Explainable machine learning approach to yield and quality improvements using deep topological data analytics
November 20, 2023
Root Cause Analysis enhanced by Causal Discovery and Topological Data Analysis
We present a novel approach to causal discovery using boosting trees together with TDA
September 14, 2023
Why you should use Topological Data Analysis over t-SNE or UMAP?
We compare the results generated from TDA with results from t-SNE and UMAP packages
April 17, 2023
What 6.5 million of #coronavirus tweets reveal about people thoughts during the pandemic
We apply Topological Data Analysis and Deep Learning to a large volume of text data to reveal hidden patterns in discussions
June 15, 2020
Using topological text analysis for COVID-19 Open Research Challenge
My take on COVID-19 Kaggle challenge analysis of scientific white papers. This research is a first step to help specialist in virusology, pharmacy and microbiology to find answers to the problem
March 24, 2020
AI for AI (artificial insemination) — Deep Topological Analysis for sensors
Find out how TDA and Deep Learning classifies events and discovers hidden patterns in IoT activity sensors’ data streams
March 10, 2020

Use Cases

User activity segmentation

DataRefiner allows you to visualise complex user activities and to find activity patterns that are important for your business. Group your users and tailor your app or a website to their needs.
Example: Find “personas” in your activity data
Add to A/B test

User Activities

Activities, captured from user interactions with the app

S3, RedShift or in-house DBs

Activities are stored in the database or a message queue


DataRefiner performs segmentation of user activities and extracts users' personas

Export segmentation result

Download a user segment using DataRefiner API

Perform a targeted marketing campaign

Target user groups precisely according to their activity patterns and be GDPR compliant

Analysis of text or user comments

Perform segmentation of the posts, user comments or any other user generated content. Fight spammers, scammers and other unwanted behaviour on your platform. Using DataRefiner you're able to visualise and understand common patterns much faster than with traditional NLP approach.
Example: Fight spammers and scammers by analysing your users' text data

Collection of texts or messages

User comments or posts, stored in your databases

CSV file with the texts

Prepare a collection of comments or user posts as CSV file


DataRefiner performs segmentation of texts and extracts common topics

A data analyst reviews the structure

An analyst understands the formed clusters and extracts the anomalies

Present the findings

Use DataRefiner player to show your findings to internal or external audiences

Analysis of sensors data (IoT) and
anomaly detection

Analyse telemetry from your hardware equipment: civil aerospace, defence, power systems, robotics and any other source of complex IoT data. Find the unexpected operational patterns and improve the performance.
Example: Find the unexpected operational patterns in aircraft engines

Time series data streams from sensors

A stream of real-time data from sensors

AWS or Azure cloud infrastructure

Raw sensor data is stored and pre-processed


DataRefiner performs segmentation of telemetry data

A data analyst reviews the structure

A subject matter expert reviews the structure and identifies the unexpected patterns

Present the findings

The findings are presented to equipment engineers for further review and re-configuration

Analysis of the bank or blockchain transactions

Identify patterns in the bank or blockchain transactions. Visualise complex schemes: fight money laundering and fraud. Understand how your initial coin offering performs and identify real users and bots among your clients.
Example: Find money laundering schemes

Crypto currency ledger

A list of account transactions with related metadata

Bank account ledger with metadata

A list of account transactions with related metadata


DataRefiner finds account activity patterns

Data Analyst / Regulator reviews the structure

Data structure reveals different account patterns including undesirable ones

Bank security review the threat

A list of suspicious accounts is reviewed in details

Analysis of graph data for ratings,
votes or questionnaires

Analyse a wide range of graph and network data: social graph, user activity graph, reviews, telecom network statuses. Visualise and understand interaction and differences between the subgroups.
Example: Find the different groups of audiences using their ratings or likes on your platform

Collection of user ratings or answers

User rating for movies, music, cafes; user votes or answers to multiple selection questionnaires

CSV file with the votes

Collection of user votes and ratings in one CSV file for analysis


DataRefiner finds common patterns and presents its structure

A data analyst describes user groups

An analyst reviews users grouped by their preferences and identifies the key questions or important issues for a community

Present the structure to a monitoring team or a client

Use DataRefiner player shows your findings to the internal or external audiences

Cyber security (analysis of the packets and flows)

Identify activity patterns in your network using data from a wide range of agents. Allow your security analysts to identify the dynamic threats by going beyond the predefined security patterns.
Example: Identify the dynamic threats in your corporate network
Agent 1
Agent 2
Agent 3
Agent X
Reports on previously suspected activity
Update the rules

Network monitoring agents

A large number of network monitoring agents are performing flows and packets monitoring

Aggregated signals from agents

The stream of signals from monitoring agents


DataRefiner constantly learns on a data stream and produces activity patterns

A data analyst reports suspicious patterns

An analyst identifies suspicious network behaviour and marks it

Cyber security system

DataRefiner reports finding to an internal cyber security system used by security analysts.

Let’s get in touch

Receive updates

Please input your contacts if you wish to receive DataRefiner updates.

Be the first to get access

Sign up to be the first who receive an invite to a closed beta. Please share your opinion, your experience is very important to us.

We're here to help

We work closely with you to find out how DataRefiner may meet the needs of your organisation. Provide details about what you plan to achieve and we will contact you as soon as possible.