Product Forecasting for a Agriculture Company: A LightGBM model was implemented for 1000 products belonging to an agricultural company using historical data with at least 66 instances, employing lags. Forecasting was performed using Multistep for 6 months from their last sale. L2 regularization was applied, but a certain category of products still shows signs of overfitting.
Forecasting Store Sales: The goal is to forecast store sales in Ecuador, considering different store locations across the country. The hybrid model used for learning consists of two components: RandomForestRegressor and LinearRegression.
Forecasting Store Sales of Drinks The goal was to determined the daily sales for the month of Febraury 2022, to achieve this, the RandomForestRegressor model was used, which resulted in a root mean squear error RMSE of 4.65.
Stars Categorization: Classification of Star Giants or Dwarfs , consider the dataset. The categorization the Stars was studied according to criteria the Morgan–Keenan (MK) classification system. Firstly, the dataset was cleaned and then it was balanced. In order to apply three models's Machine Learning, being Random Forest Classification the one with the better results, obtaining a precision of 94% used metrics F1.
Discovering Exoplanets: Problem of Classification the Explanets. we used the source. It was worked Random Forest Classification, findings hiperparameters with RandomSearchCV, achieve a 74,65% for the metric F1. In turn, displayed Confusion Matrix
Income: The project is about determined Income for some Adults
, the target is divide in two interval, <=50K and >50K. Therefore, it is problem of Classification. The Metric used is F1 and Confusion Matrix, to obtaing a precision aproximately of 81 %.
House Prices: This project aims to implement Advance Tecnique Regression for predicting house prices. To achieve this, I have utilized a forked version of Ryan Holbrook's notebook. In this modified version, I incorporated more depth Exploratory Data Analysis. In turn, to had been added new models, fit by hypeterparameters with the library Optuna.
Emojis: In this project, the goal was determined Object Localization, correspond one Emojis. The problem consisted in dectection exactly only instance of one object. To do this, we trained a model with Convolutional Neural Network (CNN), utilized for it, the library TensorFlow. The project comes from Guided Project of Coursera.
Petal of the Metal:The current machine learning model focuses on classifying 104 types of flowers based on their images. To achieve this, new layers have been added to enhance the model's architecture. Transfer learning techniques have been employed, utilizing pre-trained applications such as Xception and DenseNet201, which serve as a foundation for the hybrid model.