patrickbrus

def get_class_activation_map(model, img):
    ''' 
    this function computes the class activation map
    
    Inputs:
        1) model (tensorflow model) : trained model
        2) img (numpy array of shape (224, 224, 3)) : input image
    '''
    
    # expand dimension to fit the image to a network accepted input size

patrickbrus

from sklearn.model_selection import train_test_split
from sklearn.experimental import enable_iterative_imputer
from sklearn.impute import IterativeImputer
# split into train and test sets
# data_df_converted is the dataframe containing all features
df_train, df_test = train_test_split(data_df_converted, test_size=0.10, random_state=42)
df_train, df_val = train_test_split(df_train, test_size=0.3, random_state=42)

# create the iterative imputer model
imputer = IterativeImputer(max_iter=20, random_state=42, verbose=1)

patrickbrus

import boto3
import sagemaker
import os


# session and role
sagemaker_session = sagemaker.Session()
role = sagemaker.get_execution_role()
# region
region = boto3.Session().region_name

patrickbrus

from sagemaker.sklearn.estimator import SKLearn

FRAMEWORK_VERSION = "0.23-1"
script_path = 'source/train_linear_regression.py'

sklearn_linear_regression = SKLearn(
    entry_point=script_path,
    framework_version=FRAMEWORK_VERSION,
    instance_type="ml.c4.xlarge",
    role=role,

patrickbrus

# Define a model 
tree = DecisionTreeRegressor(criterion="mse")

# Create the Bayesion optimization object
opt = BayesSearchCV(
    tree,
    {
        "max_depth": (5, 15),
        "splitter": ["best", "random"],
    },

patrickbrus

# Define a model 
forest = RandomForestRegressor(criterion="mse", n_jobs=n_jobs)

# Create the Bayesion optimization object
opt = BayesSearchCV(
    forest,
    {
        "max_depth": (5, 15),
        "n_estimators": (10, 50),
        "bootstrap": [True, False]

patrickbrus

from sagemaker.tuner import IntegerParameter
from sagemaker.tuner import ContinuousParameter
from sagemaker.tuner import HyperparameterTuner
from sagemaker.inputs import TrainingInput

# Define exploration boundaries (default suggested values from Amazon SageMaker Documentation)
hyperparameter_ranges = {
    'alpha': ContinuousParameter(0, 1000, scaling_type="Auto"),
    'eta': ContinuousParameter(0.1, 0.5, scaling_type='Logarithmic'),
    'gamma':ContinuousParameter(0, 5, scaling_type='Auto'),

patrickbrus

import numpy as np
import matplotlib.pyplot as plt
import pandas as pd
import seaborn as sns

# read in data from csv file
df = pd.read_csv(r"data\healthcare-dataset-stroke-data.csv")

print(df.head()) # helpful as first dive into data and features

patrickbrus

# create a list of categorical features to loop over
list_categorical_features = df.columns[(df.dtypes == np.object).values].to_list()

# create a list of numerical features to loop over
list_numerical_features = df.columns[np.logical_not((df.dtypes == np.object).values)].to_list()

# plot on histogram for each numerical feature
for numerical_feature in list_numerical_features:
    plt.title(f"Histogram of {numerical_feature}")
    plt.hist(df[numerical_feature])

patrickbrus

# create a histplot of age with hue stroke to check influence of age on class stroke
sns.histplot(data=df, x="age", hue="stroke", multiple="stack")

# create histplot of gender to check if there is any influence of gender on class stroke
sns.histplot(data=df, x="gender", hue="stroke", multiple="stack")

# create histplot of smoking_satus and check influence on class stroke 
sns.histplot(data=df, x="smoking_status", hue="stroke", multiple="stack")

# here: can be interesting to also print the percentage of stroke patients of each category