• To start right away search on Google for "Google colab", click on the first link and then click on new notebook. [Colab is a cloud based environment which provides all the resources required for model development.]

• Import the stated libraries:

  • Numpy

  • Pandas

  • Sklearn

  • matplotlib.plt

  • Seaborn

• Import Boston housing dataset from Sklearn using the following command.

  from sklearn.datasets import load_boston
  var = load_boston()
  

  A bunch object is returned by load_boston() on which we will do our further work.

• Use print statement on each element of dictionary returned by the above statement and read the results Ex: print(var.DESCR) to understand the dataset.

• Create a pandas dataframe for creating a copy of the dataset on which we will carry out further preprocessing. Use the following command:

  df = pd.DataFrame(var.data, columns=var.feature_names)
  

• Checkout references available below to know more about other arguments which can be used in DataFrame() function.

• Use functions head() and tail() to see first and last five rows of the created dataframe.

• Use describe() to get even further insights on the created data frame.

• Add another column to the dataframe and store the value of target attribute in it df['MEDV'] = var.target Confirm the addition of column using head().

• Use df.dtype or df.info to know data type various features present in the dataset. If we find categorical data, then we'll require to use different encoding methods.

• Use df.isnull().sum() to check for missing values in each column. If we find missing values, then either we will place values there or we can drop the row or column.

• Create a box plot using seaborn to see the outliers in the dataset. Generally we remove the rows having outliers from our data but for small dataset like Boston housing it can lead to a loss of a significant percentage of data.

• Create a heatmap using seaborn to find corelation between different features and labels. In model creation we will be using features having a high corelation with our target label.

• Create KDE plot of different variables using seaborn library.

• Select features for creating training and test set. In place of variable x put your features (for example, NOX which stands for Nitric Oxide content) and MEDV (Median value) as a label in y

  x1 = bdf[['NOX','RM','DIS','PTRATIO','LSTAT' ]]
  y1 = bdf['MEDV']
  

• Use train_test split() from Sklearn to create train and test sets

    x_train, x_test, y_train, y_test = train_test_split(X1, Y1, test_size =0.33,random_state = 5 )
  

• Use regression on training data

  regressor = LinearRegression()
  regressor.fit(x_train, y_train)
  y_pred = regressor.predict(x_test)
  df = pd.DataFrame({'Actual': y_test, 'Predicted': y_pred})
  df
  

from sklearn.model_selection import train_test_split 
from sklearn.linear_model import LinearRegression
from sklearn import metrics

• Display the following: Mean Absolute Error, Mean Squared Error, Root Mean Squared Error, R2 Score.

  metrics.mean_absolute_error()
  metrics.mean_squared_error()
  np.sqrt(metrics.mean_squared_error())
  metrics.r2_score()
  

The above code block is missing some arguments. Checkout the references to get a clear idea on what it's missing.

• Use array function in numpy to crate an array of target label and any one of the features. Pass this through polyfit() function. polyfit() will return the slope and intercept of regression line. Store the returned values.

  x1 = numpy.array(x_test['NOX'])
  y1 = numpy.array(y_pred)
  m,b=numpy.polyfit()
  

  The above code snippet is missing arguments for polyfit(). Checkout the references for a solution.

• Use the plot() function to plot the regression line corresponding to the chosen feature.

  plt.plot(x_test['NOX'], m*x_test['NOX'] + b)
  plt.plot(x_test['NOX'],y_pred,'o')
  plt.xlabel("Nitric Oxide Content")
  plt.ylabel("Median Value")
  

  Read about the functions and it's argument in the references section.

• Use seaborn library to create a pairplot

  sns.pairplot(bdf, x_vars=['NOX','RM','DIS','PTRATIO','LSTAT'], y_vars='MEDV', height=9, aspect=0.6, kind='reg')
  

  

  Read about the function and its arguments from references to know more about pairplot and try to play around with various argument values.