ehtashambillah Goto Github PK

acute-lymphoblastic-leukemia-cell-classification-using-bayesian-convolutional-neural-networks

In this project, we deploy the Bayesian Convolution Neural Networks (BCNN), proposed by Gal and Ghahramani [2015] to classify microscopic images of blood samples (lymphocyte cells). The data contains 260 microscopic images of cancerous and non-cancerous lymphocyte cells. We experiment with different network structures to obtain the model that return lowest error rate in classifying the images. We estimate the uncertainty for the predictions made by the models which in turn can assist a doctor in better decision making. The Stochastic Regularization Technique (SRT), popularly known as Dropout is utilized in the BCNN structure to obtain the Bayesian interpretation.

advanced-artificial-neural-network

Artificial neural network with grid search and cross validation

ann-in-caret-with-dropout

ANN model fitting in Caret

ann-with-gridsearch

Gridsearching learning rate, decay and other parameters in ANN

bayesian-machine-learning-in-breast-cancer-diagnosis

The aim of this project is to apply Bayesian Machine Learning Algorithm to predict the diagnosis condition of the patients.

classification-in-caret

Classification in Caret

customer_churn_modeling_by_xgboost

Predicting the customers that are going to leave the bank using powerful machine learning method Extreme Gradient Boosting or XGBOOST.

fatherhood-effect-in-usa-in-2016

fitting-model-with-random-grid-search-in-caret

Model fitting with random grid search

ki_core_ml

microeconometrics-project

Microeconometrics Project

model-fitting-with-grid-search-in-caret

Model fitting with grid search in Caret

official-statistics-project-tables

Statistical Tables

pandas_python

Basic pandas codes

predicting-the-age-of-abalone-sea-snails-

python_machine_learning_projects

simple-artificial-neural-networks

A simple ANN model

spark_machine_learning

sql

synthetic-minority-over-sampling-technique-smote-

For a classification problem, when classes in the dependent variable are severely imbalanced (e.g. 90 yes, 10% no), training an efficient machine learning model becomes very difficult. However with SMOTE method, we can transform the data into a balaced form and train the model efficiently.

telco-customer-churn-modeling

variable-selection-by-least-absolute-shrinkage-and-selection-operator-lasso-

All independent variables do not have the similar impact on dependent variable. Here we will try to find the independent varibles that have most significant impact on dependent variable to make the ML algorithm fast and accurate by utilizing LASSO.

variable-selection-by-recursive-feature-elimination-rfe-

All independent variables do not have the similar impact on dependent variable. Here we will try to find the independent varibles that have most significant impact on dependent variable to make the ML algorithm fast and accurate by utilizing RFE.

working-startegy