Project Details
Description
Based on the financial institutions registered on FintechSaudi (https://fintechsaudi.com/), we are collecting twitters that fulfil two requirements: at least the name of one F...Financial Institution (FI) appears on the body and the twit is written in English. Two types of analysis are conducted on this big textual dataset. The quantitative Content Analysis is a discipline aimed to gather, organize, and investigate complex data as audio, image or text. Framed inside the CA, we apply two statistical methodologies. A type of unsupervised machine learning algorithm named Latent Dirichlet Allocation (LDA) is used to classifying the documents, in an analogous way as the well-known cluster analysis. This classification allows us to identify the main topics abound the investigated financial institutions, and consequently to create groups of documents based on contents similarities. Then a Latent Semantic Analysis (LDA) allow us to disclose characteristics words, the chronological evolving of the vocabulary and future trends for the Saudi financial system. The second part comprises a Sentiment Analysis (SA), which is aimed to identify and extract subjective information from our textual dataset. SA assists decision makers to get insights about social sentiments related business performance, brand perception or customers experiences, among others. Based on neural networks specifically trained for these purposes, sentiment analysis assigns labels to each document, given a confidence score. A document is labelled as “negative” if the confidence score is closer to 0. On the other hand, the label is “positive” when the score is nearer to 1. The deep learning algorithm proposed by Microsoft Azure Cognitive Services, which has demonstrated to be suitable tool for conducting SA, will here applied. Among the results expected, we are grouping financial institutions according to the perceived social sentiment. The above will allow us to disclose main differences between institutions. Finally, an ideal profile will be proposed based on those institutions associated to higher confidence scores.
Status | Finished |
---|---|
Effective start/end date | 1/01/23 → 30/06/24 |
Fingerprint
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.