#!/usr/bin/env python
# coding: utf-8

# <a href="https://colab.research.google.com/github/Shakilgithub20/predicting_diabetes/blob/main/Diabetes_Prediction_Accuracy_94_46_ipynb.ipynb" target="_parent"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a>

# In[3]:


import numpy as np # linear algebra
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.model_selection import train_test_split
from sklearn.linear_model import LinearRegression
import sklearn
import tensorflow as tf
from tensorflow import keras
from keras.models import Sequential
from keras.layers import Dense
from sklearn.metrics import accuracy_score


# In[4]:


dataset = pd.read_csv('/content/diabetes.csv')


# In[5]:


dataset.head()#viewing the head of the dataset


# In[6]:


dataset.info(verbose=True)


# In[7]:


#finding the shape of the dataset
dataset.shape


# In[8]:


#finding out the null values of the dataset
dataset.isnull().sum()


# In[9]:


f, ax = plt.subplots(1, 2, figsize = (12, 6))
f.suptitle("Diabetes?", fontsize = 18.)
_ = dataset.Outcome.value_counts().plot.bar(ax = ax[0], rot = 0, 
                                            color = (sns.color_palette()[0], sns.color_palette()[2])).set(xticklabels = ["No", "Yes"])
_ = dataset.Outcome.value_counts().plot.pie(labels = ("No", "Yes"), autopct = "%.2f%%", 
                                            label = "", fontsize = 13., ax = ax[1],\
colors = (sns.color_palette()[0], sns.color_palette()[2]), wedgeprops = {"linewidth": 1.5, "edgecolor": "#F7F7F7"}), 
ax[1].texts[1].set_color("#F7F7F7"), ax[1].texts[3].set_color("#F7F7F7")


# In[10]:


#this is also a one way to find the missing values in the dataset
dataset.describe()


# In[11]:


dataset.dtypes


# In[12]:


plt.style.use('classic')
plot = dataset.hist(figsize = (20,20))


# In[13]:


#Scatter matrix of uncleaned data
sns.pairplot(dataset )


# In[14]:


#Pair plot for clean data
sns.pairplot(data=dataset,hue='Outcome',diag_kind='kde', kind="reg")
plt.show()


# In[15]:


plt.figure(figsize = [10, 10])
sns.heatmap(dataset.corr(), annot = True, fmt = '.3f', cmap = 'vlag_r', center = 0);


# In[16]:


print(np.array(dataset))


# In[17]:


# drop columns number 9
train_data_x = dataset.drop(columns = 'Outcome', axis = 1)
train_data_y = dataset['Outcome']


# In[18]:


train_Data, test_Data, train_Out, test_Out = train_test_split(train_data_x,train_data_y,test_size=0.2, random_state=10)
print(train_Out)


# In[19]:


data_model = Sequential()
data_model.add(Dense(100, input_dim=8, activation='sigmoid'))
data_model.add(Dense(50, activation='sigmoid'))
data_model.add(Dense(30, activation='sigmoid'))
data_model.add(Dense(10, activation='sigmoid'))
data_model.add(Dense(1, activation='sigmoid'))

data_model.compile(loss='MSE', optimizer='adam', metrics=['accuracy'])
print(data_model.summary())


# In[20]:


# Compile model
# using opimizer Adam
# using loss binary_crossentropy
# data_model.compile(optimizer=tf.keras.optimizers.Adam(), loss=tf.keras.losses.binary_crossentropy, metrics=['accuracy'])
data_model.compile(optimizer='Adam', loss="binary_crossentropy", metrics=['accuracy'])


# In[21]:


hist = data_model.fit(train_Data, train_Out, validation_data=(test_Data, test_Out), epochs=1000, verbose=1)


# In[22]:


# Train and Test accuracy
scores = data_model.evaluate(train_Data,train_Out)
print("Training Accuracy: %.2f%%\n" % (scores[1]*100))
scores = data_model.evaluate(test_Data,test_Out)
print("Testing Accuracy: %.2f%%\n" % (scores[1]*100))


# In[23]:


# Final Prediction
# print(data_model.predict(test_Data))


# In[24]:


y_pred = data_model.predict(test_Data)
y_pred = (y_pred > 0.5)
print(data_model.predict(test_Data))


# In[25]:


acc = hist.history['accuracy']
loss = hist.history['loss']
plt.plot(hist.history['accuracy'])
plt.title('model accuracy')
plt.ylabel('accuracy')
plt.xlabel('epoch')
plt.legend(['train'], loc='upper left')
plt.show()
plt.savefig('accuracy_curve.png')


# In[26]:


plt.plot(hist.history['loss'])
plt.title('model loss')
plt.ylabel('loss')
plt.xlabel('epoch')
plt.legend(['train'], loc='upper left')
plt.show()
plt.savefig('loss_curve.png')