Perspectives On Data-Driven failure diagnosis : With a case study on failure diagnosis at an Payment Service Provider
Martinez-Delgado Rubio, Jose (2021-07-05)
Perspectives On Data-Driven failure diagnosis : With a case study on failure diagnosis at an Payment Service Provider
Martinez-Delgado Rubio, Jose
(05.07.2021)
Julkaisu on tekijänoikeussäännösten alainen. Teosta voi lukea ja tulostaa henkilökohtaista käyttöä varten. Käyttö kaupallisiin tarkoituksiin on kielletty.
avoin
Julkaisun pysyvä osoite on:
https://urn.fi/URN:NBN:fi-fe2021080642362
https://urn.fi/URN:NBN:fi-fe2021080642362
Tiivistelmä
Data-driven failure diagnosis aims to extract relevant information from a dataset in an automatic way. In this paper it is being proposed a data driven model for classifying the transactions of a Payment Service Provider based on relevant shared characteristics that would provide the business users relevant insights about the data analyzed.
The proposed solution aims to mimic processes applied in industrial organizations. However, the methods discussed in this paper from these organizations does not directly deal with the human component in information systems. Therefore, the proposed solution aims to offer the relevant error paths to help the business users in their daily tasks while dealing with the human factor in IT systems. The built artifact follow the next set of steps:
• Categorization of variables following data mining techniques.
• Assignation of importance for variables affecting the transaction process using predictive machine learning method.
• Classification of transactions in groups with similar characteristics.
The solution developed effectively and consistently classify more than 90% of the faults in the database by grouping them in paths with shared characteristics and with a relevant failure rate. The artifact does not depends in any predefined fault distribution and satisfactorily deal with highly correlated input variables. Therefore, the artifact has a scalable potential if previously, a data mining categorization of variables is performed. Specially, in companies that deals with rigid processes.
The proposed solution aims to mimic processes applied in industrial organizations. However, the methods discussed in this paper from these organizations does not directly deal with the human component in information systems. Therefore, the proposed solution aims to offer the relevant error paths to help the business users in their daily tasks while dealing with the human factor in IT systems. The built artifact follow the next set of steps:
• Categorization of variables following data mining techniques.
• Assignation of importance for variables affecting the transaction process using predictive machine learning method.
• Classification of transactions in groups with similar characteristics.
The solution developed effectively and consistently classify more than 90% of the faults in the database by grouping them in paths with shared characteristics and with a relevant failure rate. The artifact does not depends in any predefined fault distribution and satisfactorily deal with highly correlated input variables. Therefore, the artifact has a scalable potential if previously, a data mining categorization of variables is performed. Specially, in companies that deals with rigid processes.