This project is the final team project for the 2019 Digital Talent Scholarship with the theme Machine Learning. We used Fashion MNIST as the processed dataset in CNN modeling.

Fashion MNIST is a dataset belonging to Tensorflow. The dataset has been divided into training data, amounting to 60000, and testing data, amounting to 10000. I will display an example image of the data below.

This project features classification modeling with big data, so the neural network model is considered a suitable solution. We implement a neural network in the form of CNN. This is because CNN is quite good as a learning model with data in the form of images. This image below display how our model represented. The model we built 1-convolutional layer of CNN with 100 epochs, 128 batch size, learning rate is 0.001, and use adam optimizer.

After we do the learning and testing, we can make a conclusion that, for Fashion MNIST dataset, the accuracy score we get is 98.38% for the training data and 92.78% for the testing data.

We wanted to know more about how well our model predict the data class. You can see the result in the form of precision-recall from classification report for each class. Besides that, we presented the number of wrong prediction from our CNN model. The documentations are showed below. Based on the pictures, we can see that our model predict data from Class 1 mostly accurate than others. So that, the precision & recall have highest values.

Classification Report	Wrong Prediction

Feels free to contacts me on linkedin!