We want to use the layers of Keras to process our data of air quality In this GitHub repository, our focus is on preparing the data for analysis. We begin by normalizing the data to ensure consistency and eliminate any variations in scale. This step is crucial for enhancing the performance of our machine learning models.
After data normalization, we proceed to address the presence of outliers. Outliers can adversely affect the accuracy and reliability of our analysis, so we implement techniques to identify and remove them. By eliminating outliers, we aim to create a more robust and representative dataset for training our models.
Once the data is properly prepared, we move on to the core of our work—building and training our machine learning model. We leverage neural networks as the foundation for our model and fine-tune it using appropriate loss functions to minimize errors and improve accuracy. By optimizing these parameters, we aim to achieve the best possible performance from our neural network model.
To assess the effectiveness of our neural network, we compare its performance with other machine learning algorithms. This comparative analysis allows us to evaluate the strengths and weaknesses of different approaches and determine which algorithm yields the most accurate predictions for our specific dataset.
Through this repository, we aim to provide a comprehensive framework for data preparation, model training, and performance evaluation in machine learning. By following these steps, researchers and practitioners can gain insights into the effectiveness of various algorithms and make informed decisions when working with their own datasets. ! pip install -q kaggle
import pandas as pd df = pd.read_csv('AQI.csv')
df
print(len(df) , df.shape)
print(df.info())
df.drop(columns = ['Id' ,'Mounths'], inplace=True, axis=1)
print(len(df) , df.shape , df.info())
df.head(5)
df.isnull().sum()
import seaborn as sns sns.displot(df['PM10 in æg/m3'])
sns.displot(df['SO2 in æg/m3'])
sns.distplot(df['NOx in æg/m3'])
sns.displot(df.AQI)
df['PM10 in æg/m3'].fillna(df['PM10 in æg/m3'].mean(),inplace = True) df['SO2 in æg/m3'].fillna(df['SO2 in æg/m3'].mean(),inplace = True) df['NOx in æg/m3'].fillna(df['NOx in æg/m3'].mean(),inplace = True) df['AQI'].fillna(df['AQI'].mean(),inplace = True)
df.isnull().sum()
sns.boxplot(df.AQI)
q1 = df.AQI.quantile(.25) q3 = df.AQI.quantile(.75) IQR = q3 -q1 IQR
upper_limit = q3 + 1.5 IQR lower_limit = q1 - 1.5 IQR
df.median()
import numpy as np df['AQI'] = np.where(df['AQI'] > upper_limit , 104 , df['AQI'])
sns.boxplot(df.AQI)
x = df.drop(columns = ['AQI'] , axis = 1) x.head()
y = df.AQI y.head()
from sklearn.model_selection import train_test_split xtrain , xtest , ytrain , ytest = train_test_split(x , y , test_size = .3 , random_state = 42)
xtrain.shape , xtest.shape , ytrain.shape , ytest.shape
x_val = xtrain[:15] partial_x_train = xtrain[15:] y_val = ytrain[:15] partial_y_train = ytrain[15:]
"""# Go Ahead"""
from sklearn.linear_model import LinearRegression lm = LinearRegression() lm.fit(xtrain , ytrain)
y_pred = lm.predict(xtest)
from sklearn.metrics import mean_squared_error mse = mean_squared_error(ytest , y_pred) print('Mean Squared Error:' , mse)
import numpy as np from sklearn.metrics import mean_squared_error mse = mean_squared_error(ytest , y_pred) rmse = np.sqrt(mse) print('Root Mean Squared Error:' , rmse) from sklearn.metrics import mean_absolute_error mae = mean_absolute_error(ytest , y_pred) print('Mean Absolute Error' , mae)
"""# Neural Network"""
from tensorflow.keras.models import Sequential from tensorflow.keras.layers import Dense
reg_model = Sequential() reg_model.add(Dense(7 , activation = 'relu')) reg_model.add(Dense(256, activation = 'relu')) reg_model.add(Dense(64, activation = 'relu')) reg_model.add(Dense(32, activation = 'relu')) reg_model.add(Dense(16, activation = 'relu')) reg_model.add(Dense(1, activation = 'softmax'))
reg_model.compile(optimizer="adam",loss="mse",metrics=['mse','mae']) reg_model.fit(xtrain,ytrain,epochs=10,batch_size=2,validation_data=(xtest,ytest))
y_pred = reg_model.predict(xtest)
from tensorflow import keras
from sklearn.metrics import mean_squared_error mse = mean_squared_error(ytest, y_pred) print('Mean Squared Error:', mse)
import numpy as np from sklearn.metrics import mean_squared_error mse = mean_squared_error(ytest, y_pred) rmse = np.sqrt(mse) print('Root Mean Squared Error:', rmse)
from sklearn.metrics import mean_absolute_error mae = mean_absolute_error(ytest, y_pred) print('Mean Absolute Error:', mae)
corr_matrix = df.corr() plt.figure(figsize = ( 10 , 8)) sns.heatmap(corr_matrix , annot = True , cmap = 'coolwarm' , linewidths = .5) plt.title('Correlation Matrix') plt.show()
"""# Random Forest"""
from sklearn.ensemble import RandomForestRegressor rf = RandomForestRegressor() rf.fit(xtrain , ytrain) y_pred = rf.predict(xtest) from sklearn.metrics import mean_squared_error mse = mean_squared_error(ytest , y_pred) print('Mean Squared Error:' , mse)
import numpy as np from sklearn.metrics import mean_squared_error mse_2 = mean_squared_error(ytest , y_pred) rmse_2 = np.sqrt(mse_2) print('Root Mean Squared Error:' , rmse)
from sklearn.metrics import mean_absolute_error mae = mean_absolute_error(ytest , y_pred) print('Mean Absolute Error:' , mae)
"""# Decision Tree"""
from sklearn.tree import DecisionTreeRegressor dt = DecisionTreeRegressor() dt.fit(xtrain , ytrain) y_pred = dt.predict(xtest) from sklearn.metrics import mean_squared_error mse_3 = mean_squared_error(ytest, y_pred) print('Mean Squared Error:', mse)
import numpy as np from sklearn.metrics import mean_squared_error mse_3= mean_squared_error(ytest, y_pred) rmse_3 = np.sqrt(mse) print('Root Mean Squared Error:', rmse)
from sklearn.metrics import mean_absolute_error mae_3 = mean_absolute_error(ytest, y_pred) print('Mean Absolute Error:', mae)