| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 11.27 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
O presente documento tem como principal objetivo documentar todo o trabalho desenvolvido durante o projeto realizado na empresa Flipkick.
A assistência em eventos desportivos é um aspeto bastante importante para os clubes. Se estes conseguirem prever se os seus adeptos vão assistir ao próximo jogo seria uma informação preciosa, da forma que poderão trabalhar sobre estes adeptos, através de estratégias de marketing de modo a cativá-los para os seus estádios.
Neste contexto, surge a oportunidade de desenvolver uma aplicação capaz de efetuar tal previsão. Passando assim ao estudo das temáticas necessárias para o desenvolvimento e posteriormente implementar.
Para trabalhar com grandes quantidades de informação, é necessário recorrer a metodologias que nos permitam extrair informação, como a Aprendizagem Automática que permite realizar cálculos matemáticos de forma a realizar previsões de informação futura.
Os testes computacionais realizados demonstram que os fatores considerados para a execução de cálculos do algoritmo tem níveis de acerto mais elevados, de acordo com a quantidade utilizada para o efeito, revelando assim que mais fatores apresentam previsões mais acertadas aproximando-se mais da realidade. O algoritmo que mais consistência demonstrou foi o Two-Class Support Vector Machine, apresentando percentagens de acerto mais elevadas, não dependendo este dos fatores utilizados.
A capacidade de realizar previsões do trabalho desenvolvido irá contribuir para que os clubes aumentem a assistência nos seus estádios, através da informação extraída pelo modelo de Machine Learning desenvolvido.
The main goal of this document is documenting all developed work during the project, realized in the company Flipkick. Attendance at sporting events is a very important aspect for clubs. If they can forecast when their fans will attend the next game it would be a precious information so they can work on these supporters captivate them for their stadiums through marketing strategies. In this context the opportunity arises to develop an application capable of making such forecast, starting study of necessary themes for development and later implement. In order to work with large amounts of information, it is necessary to resort some methodologies that allow us to extract information such as Machine Learning that allows to perform mathematical calculations in order to make forecasts of future information. The computational tests carried out show that the factors considered for the execution of algorithm calculations have higher levels of success rate according to the quantity used for the effect revealing that more factors present better predictions approaching to the reality. The algorithm that presented the highest consistency was the Two-Class Support Vector Machine, presenting higher percentages of accuracy, not depending on the factors used. The ability to carry out forecasts of this work will help clubs increase attendance at their stadiums through information derived from the developed Machine Learning model.
The main goal of this document is documenting all developed work during the project, realized in the company Flipkick. Attendance at sporting events is a very important aspect for clubs. If they can forecast when their fans will attend the next game it would be a precious information so they can work on these supporters captivate them for their stadiums through marketing strategies. In this context the opportunity arises to develop an application capable of making such forecast, starting study of necessary themes for development and later implement. In order to work with large amounts of information, it is necessary to resort some methodologies that allow us to extract information such as Machine Learning that allows to perform mathematical calculations in order to make forecasts of future information. The computational tests carried out show that the factors considered for the execution of algorithm calculations have higher levels of success rate according to the quantity used for the effect revealing that more factors present better predictions approaching to the reality. The algorithm that presented the highest consistency was the Two-Class Support Vector Machine, presenting higher percentages of accuracy, not depending on the factors used. The ability to carry out forecasts of this work will help clubs increase attendance at their stadiums through information derived from the developed Machine Learning model.
Description
Keywords
Aprendizagem Automática Previsão Percentagem de acerto Machine Learning Forecast Accuracy