| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 1.84 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
Atualmente, é cada vez mais evidente e visível a quantidade de dados, que crescem a um ritmo exponencial dentro das organizações e, como tal, existe a necessidade de implementação de ferramentas de Business Intelligence (BI), capazes de fazerem o seu tratamento e posterior extração de informação e conhecimento, que permitam facilitar e dar suporte ao processo de tomada de decisão. De forma complementar, surge o Business Analytics, que, aliado ao BI se traduz numa reforçada capacidade de análise e criação de valor para o negócio, que neste caso em concreto, se irá refletir na exploração da componente Predictive Analytics baseada em modelos de séries temporais. O objetivo desta investigação foi o de aferir e identificar o potencial que as funcionalidades da ferramenta Power BI Desktop, na sua versão mais elementar, contribuem para um contexto de análise preditiva direcionada para utilizadores com conhecimentos superficiais de matérias de Data Science. Foi, assim, concebido um tutorial explicativo e formativo das etapas imprescindíveis à realização de uma análise preditiva associado a um estudo de investigação exploratória, com testes aos diversos parâmetros de configuração, influenciadores da qualidade da mesma. Para este propósito, adotou-se o método Knowledge Discovery in Databases (KDD) no que respeita ao tratamento de dados e partiu-se do modelo Selecting, Organizing, Integrating (SOI) como base para a elaboração do tutorial. A análise realizada permitiu verificar que existem três parâmetros essenciais, com influência direta no desempenho do algoritmo de previsão e consequente qualidade da mesma, sendo estes o ignorar último, o intervalo de confiança e a sazonalidade.
Nowadays, it is increasingly evident and visible the amount of data, which grows at an exponential pace within organizations and, as such, there is a need to implement Business Intelligence (BI) tools, capable of making their treatment and later extraction of information and knowledge, to facilitate and support the decision-making process. In a complementary way, Business Analytics emerges, which, combined with BI, translates into an enhanced capacity for analysis and value creation for the business, that in this specific case, will be reflected in the exploration of the Predictive Analytics component based on time series models. The purpose of this investigation was to assess and identify the potential of Power BI Desktop’s functionalities, in its most elementary version, contribute to a context of predictive analysis aimed at users with superficial knowledge of Data Science subjects. Thus, an explanatory and formative tutorial of the essential steps to carry out a predictive analysis associated with an exploratory research study was designed, with tests on the various configuration parameters, influencing the quality of the prediction. For this purpose, the Knowledge Discovery in Databases (KDD) method was implemented regarding data treatment and the selecting, organizing, integrating (SOI) model was used as the basis for the elaboration of the tutorial. The analysis made it possible to verify that there are three essential parameters, with direct influence on the performance of the forecasting algorithm and its consequent quality, such the ignore last, the confidence interval and seasonality.
Nowadays, it is increasingly evident and visible the amount of data, which grows at an exponential pace within organizations and, as such, there is a need to implement Business Intelligence (BI) tools, capable of making their treatment and later extraction of information and knowledge, to facilitate and support the decision-making process. In a complementary way, Business Analytics emerges, which, combined with BI, translates into an enhanced capacity for analysis and value creation for the business, that in this specific case, will be reflected in the exploration of the Predictive Analytics component based on time series models. The purpose of this investigation was to assess and identify the potential of Power BI Desktop’s functionalities, in its most elementary version, contribute to a context of predictive analysis aimed at users with superficial knowledge of Data Science subjects. Thus, an explanatory and formative tutorial of the essential steps to carry out a predictive analysis associated with an exploratory research study was designed, with tests on the various configuration parameters, influencing the quality of the prediction. For this purpose, the Knowledge Discovery in Databases (KDD) method was implemented regarding data treatment and the selecting, organizing, integrating (SOI) model was used as the basis for the elaboration of the tutorial. The analysis made it possible to verify that there are three essential parameters, with direct influence on the performance of the forecasting algorithm and its consequent quality, such the ignore last, the confidence interval and seasonality.
Description
Keywords
Business analytics Business intelligence Power BI Análise preditiva Predictive analytics