| dc.description.abstract | One of the most contributing parts of porosity is the presence of fractures or fractures. Fracture porosity can be estimated specifically by using log data such as density, neutron porosity, transit time/sonic, as well as mud characteristics (fluid density and transit time on fluid saturation). If one of the data that becomes the parameter does not exist, then the calculation using the fracture porosity equation approach cannot be carried out. This study aims to solve the problem of determining the fracture porosity value at a certain depth interval whose value cannot be determined because it does not have sufficient data based on the equation approach used.
Machine learning algorithms used in this research methodology, among the algorithms used are Linear Regression, Random Forest, Adaptive Boosting, Gradient Boosting, and Artificial Neural Network. The algorithm is proven in several examples of problems to make predictions. It is hoped that this algorithm can predict fracture porosity with existing data. The process of predicting fracture porosity at a certain depth is assisted by conventional log data such as Gamma Ray, Resistivity, and others which are used as a test or reference data. The results of the fracture porosity based on each test data were analyzed for similarities with the fracture porosity values generated through an equation approach.
The results of this study can be seen that some of the data used as synthetic values are quite similar to the results of the fracture porosity equation, so it can be concluded that this method is quite good in predicting fracture porosity with certain data and cases. These results are expected to be taken into consideration in solving problems if there is a lack of data, so that interpretation through log data can still be done properly | en_US |
