ML_testing

Repository for investigating sklearn and tensorflow methods based on online data.

Progress theoretical application

Tracking progress of completly self-written ML algorithms.

lin_regression_manual.ipynb

• Input data with pandas
• Investigate data with DataWrangler
• def regression_fcn: Use linear Regression model
• def loss_function: Use square root function for loss function
• def gradient: Calculating gradients manually
• main: apply functions and iterate to local optimum with def gradient

Progresss with sklearn

Tracking the progress in different Python scripts.

lin_regression_sklearn.ipynb

• Input data with pandas
• Investigate data with DataWrangler
• Shuffle data with ShuffleSplit
• Apply Linear Regression Model of sklearn
• Determine best model of cross validation with validation set
• Plot linear Model with test data
• Calculate correlation coefficients
• Estimate p-values for probability test and significance test
• Apply to another data set with KFold

neuronalnetwork_sklearn.ipynb

• Input data with pandas
• Investigate data with DataWrangler
• Apply MLPClassifier
• Calculate ML metrices

sklearn.ipynb

• Input official data of breat cancer with pandas
• Investigate data with DataWrangler
• Plot data
• Split data in train, valid and test sets
• Use LogisticRegression, DecisionTreeClassifier or RandomForestClassifier to classify breast cancer
• Determine ML metrices

Progress with tensorflow

Tracking the progress in different Python scripts.

compare_sklearn_tensorflow.ipynb

• Input official data of titanic dataset
• Cleaning dataset
• Plotting variance plots with seaborn
• Replacing string sex data with binary data
• Calculating linear correlation coeffient
• Create new feature relatives
• Select, split and scale data
• Use naiveBayes model GaussianNB
• Calculate ML metrices
• Import tensorflow 3-layer Deep-Dense-Model with ReLu, ReLu and Sigmoid activation functions
• Use ADAMS optimizer
• Apply Tensorboard
• Fit model and print ML metrices
• Compare to sklearn results