Forecasting is a method to predict a timeline dataset, but the presence of multi-seasonality can make an ambiguous result. There are various ways to handle this, and in this project, I'm using MTLS to break the multi-seasonality pattern so the model can perform better.

To make a prediction, I made 3 different models of machine learning, they are Naive Bayes, Decision Tree, and Random Forest. As a final result, RF has the best performance among other models, it has 96,6% accuracy using train data and 96,8% using test data (not overfit). Also, using the ROC plot, it almost has a perfect elbow curve (has a high proportion of true negative to the proportion of false negative (or the model has high true positive and true negative), and 99,9% AUC (the model can separate each class almost perfectly)

I'm using GLM to predict the survival probability of Titanic passengers. As the final result, the model has 74% accuracy and 86% sensitivity. As a comparison, I also try to build a model using K-Nearest Neighbor and compare its performance with GLM's final model.

This is a simple example usage of linear regression to predict house prices. The main idea of this analysis is to explain how step-by-step doing linear regression.

Here you can see the architecture used in HomeScape application. Whatever is done behind HomeScape, starting from read data, machine learning building, choropleth map building until the plot I used in the application is explained here.

This project was created as a requirement to fulfill the second capstone project of machine learning material. The conclusion obtained is the best model to handle multiseasonal using STLM with the ARIMA method, and the final MAE value obtained is 4.82.

This analysis aims to get a better understanding of customer behavior in buying terms of deposits. This analysis is essential for banks or any financial institution to know before they contact their customer.