Machine Learning to measure reputational risk
The tool developed by nfqSolutions allows a financial institution to measure the reputational risk. An algorithm, once being trained by using machine learning techniques, catalogues the feeling of the Twitter entries related to the referred financial institution.
How can a bank know what the status of its reputation is?
How can you predict a reputational crisis?
Is it possible to take preventive measures to avoid a potentially damaging reputational spiral to the institution?
The concept of Reputational Risk
The financial market is based on an ongoing state of trust between its stakeholders. For this, the reputation of an entity is a key factor to consider when referring to the stability of its assets as well as a major source of risk when dealing with potential crises.
As such it is recognized by the European Banking Authority (EBA), which integrates reputational risk measurement within its risk control requirements for banking entities.
Qdos uses Twitter to measure the status of an entity´s reputation
The tool tracks and stores bank statements and performs a sentiment analysis (positive, negative, neutral) conductive to obtain a reliable picture of the entity’s reputational status over a period of time.
In addition, through a continued study of the entity’s reputational behaviour, Qdos can predict the risk of a future reputational crisis and further spirals that could lead to different risk situations for the entity.
Qdos tracks and stores bank statements and performs a sentiment analysis (positive, negative, neutral) to obtain a reliable picture of the entity’s reputational status over time.
Twitter is considered as the ideal social platform in order to study a company’s reputation:
- The character limitation of a tweet allows a more structured analysis of the information.
- The ability to subscribe to third-party content (followers), retweets and hashtags makes it a very reliable network for measuring the viral potential of a content.
- The richness of data that accompanies each tweet (date and time of sending, device, location, followers …) allows weighing the potential importance of the content.
- Most entities have reputation management teams on Twitter to take corrective action when needed.
Machine Learning to train a contextual algorithm
To perform this sentiment analysis nfqSolutions has developed a statistical algorithm capable of assigning characteristics (“feelings”) to those tweets previously identified as relevant for the entity.
The algorithm is trained using principles of supervised learning with a dynamic set that is being continuously updated as new tweets enter.
Social Listening and Reputational Risk
Qdos performs a double analysis of the collected tweets to be able to measure more than just the current status of an entity’s reputation:
1. Historical analysis: detection of:
- Behavioural patterns of the tweets.
- The speed of its changing evolution.
- The most influential users.
2. Online analysis: Use of the information obtained from the historical analysis to predict the possibility that a unique tweet or set of tweets can trigger or generate a spiral.
In addition, the tool has an alert manager that automatically generates warnings to the entity when existing reputational risks arise.
The power and flexibility of Qdos allows wide and extensive measurement, monitoring and management of reputational risk of an entity and its reference market.
- Reference Index of the evolution of the corporate reputation.
- Early detection tools for potential reputational crises.
- Reputational analysis of other companies within the sector and the competition.
- Quality indices of customer service.
- Integration with other risk indicators such as sanctions from official bodies.
Currently, nfq Solutions is collaborating with a major financial institution in the implementation of Qdos as a reputational risk assessment solution and its integration within other pillars of the company in order to obtain its own reputational index.
NfqSolutions is the nfq Group division specialized in the development of software solutions.
Warning: count(): Parameter must be an array or an object that implements Countable in /home/austri/public_html/wp-content/themes/nfq/single-re.php on line 273