Commit Live PR

q01_load_data/build.py

@@ -1,7 +1,14 @@
+# %load q01_load_data/build.py
 import pandas as pd
 import numpy as np
-from sklearn.model_selection import train_test_split
+
+path = 'data/elecdemand.csv'
+
+def q01_load_data(path):
+    data = pd.read_csv(path)
+    data['Datetime'] = pd.to_datetime(data['Datetime'])
+    return data.shape, data
+
 
 
 
-

q02_data_splitter/build.py

@@ -1,7 +1,18 @@
+# %load q02_data_splitter/build.py
 import pandas as pd
 import numpy as np
 from sklearn.model_selection import TimeSeriesSplit
 from greyatomlib.time_series_day_02_project.q01_load_data.build import q01_load_data
 
+def q02_data_splitter(path):
+    path = 'data/elecdemand.csv'
+    shape, df = q01_load_data(path)
+    tscv = TimeSeriesSplit(n_splits=2)
+    com_idx = []
+    for train_index, valid_index in tscv.split(df):
+        com_idx.append((train_index, valid_index))
+    return com_idx
+
+
+
 
-

q03_time_plot/build.py

@@ -1,7 +1,19 @@
+# %load q03_time_plot/build.py
 import pandas as pd
 import numpy as np
 import matplotlib.pyplot as plt
 from greyatomlib.time_series_day_02_project.q01_load_data.build import q01_load_data
 plt.switch_backend('agg')
 
+def q03_time_plot(path):
+    path = 'data/elecdemand.csv'
+    shape, df = q01_load_data(path)
+    plt.plot(df['Datetime'], df['Demand'])
+    plt.title('Electricity Demand for Australia for a year')
+    plt.xlabel('Year-Month')
+    plt.ylabel('Demand')
+    plt.show();
+
+
+
 

q04_boxplot/build.py

@@ -1,7 +1,16 @@
+# %load q04_boxplot/build.py
 import pandas as pd
 import numpy as np
 import matplotlib.pyplot as plt
 from greyatomlib.time_series_day_02_project.q01_load_data.build import q01_load_data
 plt.switch_backend('agg')
 
-
+def q04_boxplot(path):
+    path = 'data/elecdemand.csv'
+    shape, df = q01_load_data(path)
+    df.boxplot(column=['Demand'], by=['WorkDay'])
+    plt.show();
+
+
+
+

q05_feature_engineering/build.py

@@ -1,9 +1,21 @@
+# %load q05_feature_engineering/build.py
 import pandas as pd
 import numpy as np
 import matplotlib.pyplot as plt
+from scipy.stats import pearsonr
 from greyatomlib.time_series_day_02_project.q01_load_data.build import q01_load_data
 plt.switch_backend('agg')
 
+def q05_feature_engineering(path):
+    path = 'data/elecdemand.csv'
+    shape, df = q01_load_data(path)
+    corr, p_value = pearsonr(df['Temperature'], df['Demand'])
+    plt.scatter(df['Temperature'], df['Demand'])
+    plt.xlabel('Temperature')
+    plt.ylabel('Demand')
+    plt.title('Temperature vs Demand')
+    plt.show();
+
 
 
 

q05_feature_engineering_part2/build.py

@@ -1,8 +1,35 @@
+# %load q05_feature_engineering_part2/build.py
 import pandas as pd
 import numpy as np
 import matplotlib.pyplot as plt
-from greyatomlib.time_series_day_02_project.q01_load_data.build import q01_load_data 
+from greyatomlib.time_series_day_02_project.q01_load_data.build import q01_load_data
 plt.switch_backend('agg')
 
+def q05_feature_engineering_part2(path):
+    path = 'data/elecdemand.csv'
+    shape, data = q01_load_data(path)
+    data['hour'] = data['Datetime'].dt.hour
+    data['month'] = data['Datetime'].dt.month
+    plt.figure(figsize=(16, 6))
+    demand_hours = []
+    for i in range(1,25):
+        one = data[data['hour'] == i]['Demand'].values
+        demand_hours.append(one)
+    demand_months = []
+    for j in range(1,13):
+        demand_months.append(data[data['month'] == j]['Demand'].values)
+    plt.subplot(211)
+    plt.boxplot(demand_hours, labels=[str(i) for i in range(1,25)])
+    plt.xlabel('Hour')
+    plt.ylabel('Demand')
+    plt.title('Change in Electricity demand wrt to Hour') 
+    plt.subplot(212)
+    plt.boxplot(demand_months, labels=[str(i) for i in range(1,13)])
+    plt.xlabel('Months')
+    plt.ylabel('Demand')
+    plt.title('Change in Electricity demand wrt to months') 
+    plt.show();
+
+
+
 
-

q05_feature_engineering_part3/build.py

@@ -1,8 +1,33 @@
+# %load q05_feature_engineering_part3/build.py
 import pandas as pd
 import numpy as np
 import matplotlib.pyplot as plt
 from greyatomlib.time_series_day_02_project.q01_load_data.build import q01_load_data
 plt.switch_backend('agg')
 
+def q05_feature_engineering_part3(path):
+    path = 'data/elecdemand.csv'
+    shape, data = q01_load_data(path)
+    data['hour'] = data['Datetime'].dt.hour
+    data['month'] = data['Datetime'].dt.month
+    plt.figure(figsize=(16, 6))
+    demand_hours = []
+    for i in range(1,25):
+        one = data[data['hour'] == i]['Demand'].values
+        demand_hours.append(one)
+    demand_months = []
+    for j in range(1,13):
+        demand_months.append(data[data['month'] == j]['Demand'].values)
+    plt.subplot(211)
+    plt.boxplot(demand_hours, labels=[str(i) for i in range(1,25)])
+    plt.xlabel('Hour')
+    plt.ylabel('Demand')
+    plt.title('Change in Electricity demand wrt to Hour') 
+    plt.subplot(212)
+    plt.boxplot(demand_months, labels=[str(i) for i in range(1,13)])
+    plt.xlabel('Months')
+    plt.ylabel('Demand')
+    plt.title('Change in Electricity demand wrt to months') 
+    plt.show();
+
 
-

q05_feature_engineering_part4/build.py

@@ -1,9 +1,18 @@
+# %load q05_feature_engineering_part2/build.py
 import pandas as pd
 import numpy as np
-from sklearn.model_selection import train_test_split
 import matplotlib.pyplot as plt
 from greyatomlib.time_series_day_02_project.q01_load_data.build import q01_load_data
-plt.switch_backend('agg')
 
-def q05_feature_engineering_part4():    
-
+path = 'data/elecdemand.csv'
+
+def q05_feature_engineering_part4(path):
+    shape, data = q01_load_data(path)
+    data['hour'] = data['Datetime'].dt.hour
+    data['month'] = data['Datetime'].dt.strftime('%b')
+    data['Peakhours'] = list(map(lambda x: 1 if x in range(6,20) else 0, data['Datetime'].dt.hour))
+    data['Peakmonths'] = list(map(lambda x: 1 if x in ['Feb', 'May', 'Jun', 'Jul', 'Aug'] else 0, data['Datetime'].dt.strftime('%b')))
+    return data
+
+
+

q06_linear_regression/build.py

@@ -1,3 +1,4 @@
+# %load q06_linear_regression/build.py
 import pandas as pd
 import numpy as np
 import math
@@ -6,7 +7,23 @@
 from greyatomlib.time_series_day_02_project.q05_feature_engineering_part4.build import q05_feature_engineering_part4
 from greyatomlib.time_series_day_02_project.q02_data_splitter.build import q02_data_splitter
 
-fe =  ["WorkDay", "Peakhours", "Peakmonths"]
+fe =  ['WorkDay', 'Peakhours', 'Peakmonths']
+
+def q06_linear_regression(path, columns = fe, random_state = 9):
+    np.random.seed(random_state)
+    data = q05_feature_engineering_part4(path)
+    com_idx =  q02_data_splitter(path)
+    rmse = []
+    for i in com_idx:
+        train_idx = i[0]
+        valid_idx = i[1]
+        X_train, y_train = data.ix[train_idx, fe], data.ix[train_idx, 'Demand']
+        X_valid, y_valid = data.ix[valid_idx, fe], data.ix[valid_idx, 'Demand']
+        model = LinearRegression()
+        model.fit(X_train, y_train)
+        y_pred = model.predict(X_valid)
+        rms = mean_squared_error(y_valid, y_pred)**0.5
+        rmse.append(rms)
+    return np.mean(rmse)
 
 
-

q07_randomforest_regressor/build.py

@@ -1,3 +1,4 @@
+# %load q07_randomforest_regressor/build.py
 import pandas as pd
 import numpy as np
 import math
@@ -6,7 +7,23 @@
 from greyatomlib.time_series_day_02_project.q05_feature_engineering_part4.build import q05_feature_engineering_part4
 from greyatomlib.time_series_day_02_project.q02_data_splitter.build import q02_data_splitter
 
-fe =  ["WorkDay", "Peakhours", "Peakmonths"]
+fe =  ['WorkDay', 'Peakhours', 'Peakmonths']
+
+def q07_randomforest_regressor(path, columns = fe, random_state = 9):
+    np.random.seed(random_state)
+    data = q05_feature_engineering_part4(path)
+    com_idx =  q02_data_splitter(path)
+    rmse = []
+    for i in com_idx:
+        train_idx = i[0]
+        valid_idx = i[1]
+        X_train, y_train = data.ix[train_idx, fe], data.ix[train_idx, 'Demand']
+        X_valid, y_valid = data.ix[valid_idx, fe], data.ix[valid_idx, 'Demand']
+        model = RandomForestRegressor(n_estimators=50, min_samples_leaf=30, random_state=10)
+        model.fit(X_train, y_train)
+        y_pred = model.predict(X_valid)
+        rms = mean_squared_error(y_valid, y_pred)**0.5
+        rmse.append(rms)
+    return np.mean(rmse)
 
 
-

q08_gradientboosting_regressor/build.py

@@ -1,3 +1,4 @@
+# %load q08_gradientboosting_regressor/build.py
 import pandas as pd
 import numpy as np
 import math
@@ -6,5 +7,23 @@
 from greyatomlib.time_series_day_02_project.q05_feature_engineering_part4.build import q05_feature_engineering_part4
 from greyatomlib.time_series_day_02_project.q02_data_splitter.build import q02_data_splitter
 
-fe =  ["WorkDay", "Peakhours", "Peakmonths"]
+fe =  ['WorkDay', 'Peakhours', 'Peakmonths']
 
+def q08_gradientboosting_regressor(path, columns = fe, random_state = 9):
+    np.random.seed(random_state)
+    data = q05_feature_engineering_part4(path)
+    com_idx =  q02_data_splitter(path)
+    rmse = []
+    for i in com_idx:
+        train_idx = i[0]
+        valid_idx = i[1]
+        X_train, y_train = data.ix[train_idx, fe], data.ix[train_idx, 'Demand']
+        X_valid, y_valid = data.ix[valid_idx, fe], data.ix[valid_idx, 'Demand']
+        model = GradientBoostingRegressor(n_estimators=200, min_samples_leaf=10, learning_rate=0.01, random_state=random_state)
+        model.fit(X_train, y_train)
+        y_pred = model.predict(X_valid)
+        rms = mean_squared_error(y_valid, y_pred)**0.5
+        rmse.append(rms)
+    return np.mean(rmse)
+
+