Skip to content

Home

Causal Discovery
Technology that discovers hidden truths in data and identifying individual causal relationships and opens up new possibilities for business and R&D.

Challenge in Causal Discovery

Challenge in Causal Discovery
Data analysis is used in various fields, including marketing, healthcare, and materials development. However, traditional analytical methods can only capture broad trends across entire datasets, and we sometimes overlooked causal relationships that exist only under specific conditions. Even experienced professionals had difficulty discovering unexpected relationships.

Fujitsu’s Causal Discovery

Causal Discovery
Fujitsu’s “Causal Discovery” technology enables the discovery of causal relationships not only across entire datasets but also observable only under specific conditions. This will increase opportunities for unexpected discoveries and will also allow for the identification of previously overlooked, unique causal relationships. It opens up new possibilities in business and R&D.

The benefits of Fujitsu’s Causal Discovery

  1. Discovering new causal relationships
    • It discovers causal relationships that exist under specific conditions, which have been overlooked in traditional analysis. It generates new business insights for business and breakthroughs in R&D.
  2. Realization of targeted initiatives
    • It enables more effective marketing initiatives, personalized medicine, and materials development based on causal relationships specific to particular segments.
  3. Expanding expert knowledge
    • Discovering unexpected causal relationships, often missed by even experienced experts, expands their knowledge and facilitates more sophisticated decision-making.
  4. Contribution to Genomic Medicine
    • This technology will advance genomic medical research, enabling personalized medicine, including optimal treatment selection, and investigations into the origins of cancer.

Technical Overview

Target Industry/Users

Industries and users that use AI to formulate measures to solve real problems in various business fields such as healthcare and marketing.

Identify the causes of customer purchases and develop marketing measures (marketing)
Identify the causes of patients' illnesses and develop treatment plans (healthcare)

Challenges in Target Industry and Operations

Traditional data analysis has focused on estimating causal relationships within entire datasets. However, solving many practical problems requires estimating these relationships for individual data instances.

In the case of cancer treatment in the medical field, many cancer patients have been identified by the expression of unique genes that affect the disease state of cancer. In order to devise an appropriate treatment plan for individual patients, therefore, doctors must identify genes that are unique to each cancer patient, not genes that are common to all cancer patients. In the case of promotions in marketing, each customer within a larger group has a distinctive characteristic that leads to their purchase, and in order to plan appropriate outreach for individual customers, it becomes necessary to identify a characteristic, motivational cause for each customer, not a cause common to all customers.

Technical Challenges

In order to accurately estimate the characteristic cause-and-effect relationship for each data item, researchers must compare the results of different operations and actions applied to the corresponding person or object under the same conditions. However, it is difficult to obtain, for example, the results of differing gene expression in a single cancer patient or implementing different promotion measures for one customer. Therefore, the challenge was how to discover the characteristic causality of each piece of data from the entire data of various patients and customers.

Solutions

Focusing on correlations, which are less strict than causality, all groups of data with a common correlation from the entire data are extracted. In general, when a group of data has a causality, it also has a correlation. Therefore, by first identifying correlated data groups, we can efficiently exclude groups lacking significant correlations. Fujitsu has developed a technology to comprehensively identify all relevant combinations, ensuring the efficient extraction of all such data groups without omissions.

Next, we estimate causal relationships within the extracted data groups. This allows to identify all groups exhibiting characteristic causal relationships. By finding the groups that correspond to the new data for which we want to determine causality, we can infer the characteristic causal relationships within the new data.

Causal Discovery - Solutions

Fujitsu's Technological Advantage

Enables the discovery of the characteristic causal relationship of individual data.
Enables the discovery of causality specific to certain conditions, not just overall causality.

The benefits of Fujitsu’s Causal Discovery（Detailed version）

This technology will contribute to medical research in the field of genomic medicine, including personalized medicine, and help shed light on the causes of cancer.
Beyond healthcare and marketing, it can also be used in finance to access the creditworthiness of each customer, and in manufacturing to determine the cause of defects in each product.

Use Cases

This technology can be applied to various areas, including:

  • Genomic medicine: Advancing personalized medicine by tailoring treatments to individual patients, and investigating the underlying causes of cancer.
  • Manufacturing: Analyzing product defects to identify root causes and improve quality control.
  • Marketing: Analyzing customer behavior to optimize marketing campaigns and personalize customer experiences.
  • Finance: Managing financial market risk and forecasting demand.

Case studies

Technical Trial

  • A Proof of Concept (PoC) is possible.