#!/usr/bin/env python

# coding: utf-8

# In[1]:

conda install nomkl

# In[1]:

conda install seaborn

# In[101]:

import pandas as pd

import numpy as np

# In[102]:

from sklearn.linear_model import LogisticRegression

import os

os.environ['KMP_DUPLICATE_LIB_OK']='True'

# In[103]:

import seaborn as sns

# In[104]:

import matplotlib.pyplot as plt

get_ipython().run_line_magic('matplotlib', 'inline')

# In[105]:

from sklearn.linear_model import LogisticRegression

# In[106]:

import joblib

# In[107]:

import pandas as pd

diabetesDF = pd.read_csv('diabetes.csv')

# In[ ]:

# In[108]:

get_ipython().run_cell_magic('HTML', '', '<style type="text/css">\n\n table td,th {\n broder: 1px black solid !important;\n colour: black !important;\n}\n</style>')

# In[109]:

diabetesDF.head()

# # missing and zero values

# In[ ]:

# In[110]:

diabetesDF.info()

# In[111]:

diabetesDF.describe()

# In[114]:

print(diabetesDF.isnull().any().sum())

print(diabetesDF.isnull().sum())

# In[116]:

diabetesDF=diabetesDF.drop("Insulin",axis=1)

# In[117]:

diabetesDF=diabetesDF.dropna()

diabetesDF.shape

# In[118]:

diabetesDF.groupby('Outcome').mean()

# In[ ]:

# In[120]:

## to find where none of teh columns has 0 value(except the first and last column)

diabetesDF=diabetesDF[-(diabetesDF[diabetesDF.columns[1:-1]]==0).any(axis=1)]

diabetesDF.shape

# In[121]:

diabetesDF.groupby('Outcome').agg(['mean','median'])

# In[ ]:

# # Histogram plots

# In[122]:

for i,col in enumerate(diabetesDF[:-1]):

plt.figure(i)

sns.displot(diabetesDF[col]);

# ## scatter matrix

#

# In[123]:

sns.pairplot(diabetesDF, hue="Outcome", diag_kind="hist");

# ## correlation plots

# In[17]:

diabetesDF.corr()

# In[18]:

plt.figure(figsize=(9,9))

sns.heatmap(np.abs(diabetesDF.corr()), annot=True, cmap="viridis", fmt="0.2f");

# ## box plots, violin plots and bee swarm plots

#

# In[19]:

sns.boxplot(x="Outcome", y="BMI", data=diabetesDF, whis=3.0);

sns.swarmplot(x="Outcome", y="BMI", data=diabetesDF, size=2, color="k", alpha=0.3);

# In[20]:

sns.violinplot(x="Outcome", y="BMI", data=diabetesDF);

sns.swarmplot(x="Outcome", y="BMI", data=diabetesDF, size=2, color="k", alpha=0.3);

# ## 2D Histograms

# useful when you have a lot of data..i.e. at least 1000's of points

# In[21]:

plt.figure(figsize=(10,9))

plt.hist2d(diabetesDF["Glucose"],diabetesDF["BMI"], bins=(20,20),cmap="magma")

plt.xlabel("Glucose")

plt.ylabel("BMI")

plt.colorbar();

# In[22]:

## remember we do not have a large number of points

hist,x_edge,y_edge = np.histogram2d(diabetesDF["Glucose"],diabetesDF["BMI"], bins=20)

#here we want the centers of the bins

x_center=0.5 * (x_edge[1:] + x_edge[:-1])

y_center=0.5 * (y_edge[1:] + y_edge[:-1])

plt.contour(x_center, y_center, hist, levels=6)

plt.xlabel("Glucose")

plt.ylabel("BMI")

plt.colorbar();

# In[23]:

plt.figure(figsize=(9,9))

sns.kdeplot(diabetesDF["Glucose"],diabetesDF["BMI"], cmap="viridis", bw_methods=(4,4));

plt.hist2d(diabetesDF["Glucose"], diabetesDF["BMI"], bins=20, cmap="magma", alpha=0.5)

plt.colorbar();

# In[24]:

plt.figure(figsize=(9,9))

sns.kdeplot(diabetesDF["Glucose"], diabetesDF["BMI"],cmap="magma",shade=True,cbar=True);

# In[25]:

plt.figure(figsize=(9,9))

m=diabetesDF["Outcome"]==1

plt.scatter(diabetesDF.loc[m, "Glucose"],diabetesDF.loc[m,"BMI"], c="r", s=15, label="Diabetic")

plt.scatter(diabetesDF.loc[-m, "Glucose"],diabetesDF.loc[-m,"BMI"], c="b", s=15, label="Non-Diabetic")

plt.xlabel("Glucose")

plt.ylabel("BMI")

plt.legend(loc=2);

# In[26]:

pip install chainconsumer

# In[27]:

params=["Glucose","BMI"]

m=diabetesDF["Outcome"]==1

diabetic=diabetesDF.loc[m,params].values

non_diabetic=diabetesDF.loc[-m,params].values

non_diabetic.shape

# In[28]:

from chainconsumer import ChainConsumer

c= ChainConsumer()

c.add_chain(diabetic,parameters=params, name="Diabetic", kde=1.0, color="b")

c.add_chain(non_diabetic, parameters=params, name="Non Diabetic",kde=1.0, color="r")

c.configure(contour_labels="confidence",usetex=False)

c.plotter.plot(figsize=3.0);

# * It gives you two contours for each chain, one of the 68% confidence interval and the other is for the 95% confidence interval.

#

# * This is useful when doing Hypothesis testing.

#

# * For example, if you randomly pick a Diabetic person out of a different data sample, you can say that 68% of the time would expect glucose levels to lie within the 68% contour, and 95% of the time their BMI and Glucose levels would lie in the second contour.

#

# * This is useful when you would like to check if a data point comes from this distribution or not.

#

# * For example, you can check where the data point lies and estimate the chance of it being of a diabetic or Non Diabetic person.

#

# In[29]:

c.plotter.plot_summary(figsize=2.0);

# * if we ignore the correlations, we ca visually see that Diabetic individuals have higher Glucose levels that Non Diabetic individuals.

# * They also tend to have a higher BMI(look at hte tall of the curve)

#

# # A probabilistic Analysis

# In[30]:

diabetesDF=diabetesDF[["Glucose","BMI","Age", "Outcome"]]

# In[31]:

diabetesDF.shape

# In[32]:

import plotly.graph_objects as go

df_y= diabetesDF.loc[diabetesDF["Outcome"]==1,["Glucose","BMI", "Age"]]

df_n= diabetesDF.loc[diabetesDF["Outcome"]==0, ["Glucose","BMI", "Age"]]

## crete an interactive 3D plot using the two subdatasets

fig=go.Figure()

fig.add_trace(go.Scatter3d(x=df_y["Glucose"],y=df_y["BMI"], z=df_y["Age"], mode="markers", name="Diabetic"))

fig.add_trace(go.Scatter3d(x=df_n["Glucose"],y=df_y["BMI"], z=df_y["Age"], mode="markers", name="Non Diabetic"))

fig.show()

# In[33]:

sns.set(style="ticks")

sns.pairplot(df_y,diag_kind="hist");

# In[34]:

sns.set(style="ticks")

sns.pairplot(df_n,diag_kind="hist");

# In[35]:

df_y.shape

# In[36]:

df_n.shape

# In[37]:

test_point=[110,35,52]

fig=go.Figure()

fig.add_trace(go.Scatter3d(x=df_y["Glucose"],y=df_y["BMI"],z=df_y["Age"],mode="markers",name="Diabetic"))

fig.add_trace(go.Scatter3d(x=df_n["Glucose"],y=df_n["BMI"],z=df_n["Age"],mode="markers",name="Non Diabetic"))

fig.add_trace(go.Scatter3d(x=[test_point[0]],y=[test_point[1]],z=[test_point[2]],mode="markers",name="Test"))

# In[38]:

from scipy.stats import multivariate_normal as mn

prob_test=[]

for d in [df_y,df_n]:

mean=np.mean(d)

cov=np.cov(d,rowvar=0)

prob=mn.pdf(test_point,mean,cov)

prob_test.append(prob)

# In[39]:

prob_test

# In[40]:

num_y=df_y.shape[0]

num_n=df_n.shape[0]

print("Number of people with Diabetes is: ", num_y)

print("Number of people without Diabetes is: ", num_n)

prob_diagnosis=num_n * prob_test[0]/(num_y * prob_test[0] + num_n *prob_test[1])

print(f"negative diagnosis chance is { 100 * prob_diagnosis:.2f}%")

# # predicting Diabetes with logistic Regression

#

# In[124]:

diabetesDF.head()

# In[125]:

X,y= diabetesDF.values[:,:-1],diabetesDF.values[:,-1]

# In[126]:

from sklearn.model_selection import train_test_split

# In[130]:

X_train, X_test, y_train, y_test = train_test_split(X,y,test_size=0.01,random_state=42,stratify=y)

# In[131]:

model=LogisticRegression()

model.fit(X_train,y_train)

# In[132]:

accuracy=model.score(X_test,y_test)

print("accuracy", accuracy * 100, "%")

# In[135]:

coeff = list(diabetesCheck.coef_[0])

labels = list(dfTrain.drop('Outcome',1).columns)

# In[136]:

import pandas as pd

features = pd.DataFrame()

features['Features'] = labels

features['importance'] = coeff

features.sort_values(by=['importance'], ascending=True, inplace=True)

features['positive'] = features['importance'] > 0

features.set_index('Features', inplace=True)

features.importance.plot(kind='barh', figsize=(11, 6),color = features.positive.map({True: 'blue', False: 'red'}))

plt.xlabel('Importance')

# In[ ]: