These are my kaggle notebooks that I made in my spare time.
-
In this notebook, I was working on the MNIST dataset. The goal was to identify handwritten numbers close to 100 percent.
-
[Update 18.08.2020]
Creation of the CNN network -
[Update 21.08.2020]
Adding a visualization [Top 34%] -
[Update 01.09.2020]
Adding data augmentation, batch normalization and callbacks [Top 10%] -
[Update 15.09.2020]
Adding extra dataset to improve accuracy to 99.8 [Top 6%] -
[Update 03.05.2021]
Adding Occlusion sensitivity and changing metrics to f1 [Top 4%]
The final accuracy of 99.6% was achieved on the MNIST data set prepared for the [competition](https://www.kaggle.com/c/digit-recognizer) -
-
Diagnosis of pneumonia chest X-ray
I will be working with the Chest X-ray dataset (5863 JPEG) to create a universal model for pneumonia diagnosis. The goal is to know with over 90% accuracy whether a person is healthy, has bacterial or viral pneumonia.
-
[Update 02.09.2020]
Creation first version with CNN network -
[Update 08.09.2020]
Create a description
The final accuracy of ~92% was achieved. -
-
The dataset contains processed housing price data. My goal was to maximize the accuracy of price prediction for an advertisement.
- [Update 14.05.2021]
First version with EDA and modeling.
The final submission is Top 2% on leaderboard. - [Update 14.05.2021]
-
Heart Attack Analysis + prediciton 90% acc
My goal for this dataset was to analyze and predict people with heart disease.
- [Update 14.05.2021]
First version with EDA and modeling.
The final accuracy of ~90% was achieved. - [Update 14.05.2021]
- Python (Numpy, Pandas)
- Jupyter Notebook
- TensorFlow