Finished Exercise 3

Exercise 3/exercise3.py

@@ -5,7 +5,8 @@ from scipy import stats
 import pandas as pd
 from sklearn.linear_model import LinearRegression
 from sklearn.model_selection import train_test_split
-# from tqdm import tqdm #Import all needed modules
+from sklearn.linear_model import SGDRegressor
+from sklearn.preprocessing import StandardScaler
 
 columns = ["Material", "Density", "Radius", "Mass", "Temperature", "Pressure", "Height", "Time"]
 columnsNoMaterial = ["Density", "Radius", "Mass", "Temperature", "Pressure", "Height", "Time"]
@@ -62,7 +63,6 @@ def part1():
         materialDf = df[df["Material"] == material]
         for radius in radii:
             radiusDf = materialDf[materialDf["Radius"] == radius]
-            print(radiusDf)
             plt.scatter(radiusDf["Height"], radiusDf["Time"], label=f'Radius {radius}m')
 
         plt.xlabel("Drop Height/m")
@@ -76,7 +76,6 @@ def part1():
 def part2():
     dfNoMaterial = df.drop("Material", axis=1)
     corrMatrix = dfNoMaterial.corr(method='pearson')
-    print(corrMatrix)
 
     fig, ax = plt.subplots()
     im = ax.imshow(corrMatrix, cmap="gnuplot", vmin=-1, vmax=1)
@@ -127,6 +126,8 @@ def part3():
             plt.title(f'Iron data and predictions for radius of {radius}m')
             plt.show()
 
+    predict()
+    
     trainData, testData = train_test_split(ironDf, test_size=0.1)
 
     features = trainData[["Density", "Radius", "Mass", "Temperature", "Pressure", "Height"]]
@@ -146,9 +147,47 @@ def part3():
             print(f'For radius of {radius}m, the true R^2 value is {trueR2} and the predicted R^2 value is {predR2}')
             plt.show()
 
+    trueVpred()
+
     def calcResiduals():
         residualsFeatures = testData[["Density", "Radius", "Mass", "Temperature", "Pressure", "Height"]]
         residuals = testData["Time"] - trainLinearReg.predict(residualsFeatures)
 
         plt.scatter(testData["Radius"], residuals)
         plt.show()
+
+    calcResiduals()
+
+def part4():
+    reg = SGDRegressor()
+
+    regSamples = df[["Density", "Radius", "Mass", "Temperature", "Pressure", "Height"]]
+    regTargets = df[["Time"]]
+    reg.fit(regSamples, regTargets)
+    print(f'The unscaled R^2 value is: {reg.score(regSamples, regTargets)}')
+
+    scaler = StandardScaler()
+    scaledSamples = scaler.fit_transform(regSamples, regTargets)
+
+    scaledReg = SGDRegressor()
+
+    scaledReg.fit(scaledSamples, regTargets)
+    print(f'The scaled R^2 value is: {scaledReg.score(scaledSamples, regTargets)}')
+    deScaledCoefs = scaledReg.coef_ / scaler.scale_
+
+    for i in range(6):
+        print(f'The coefficient of {columns[i+1]} is {deScaledCoefs[i]} {units[i+1]}')
+
+    huberReg = SGDRegressor(loss="huber")
+
+    huberReg.fit(scaledSamples, regTargets)
+    print(f'The scaled R^2 value using the huber loss function is: {huberReg.score(scaledSamples, regTargets)}')
+    deScaledCoefs = huberReg.coef_ / scaler.scale_
+
+    for i in range(6):
+        print(f'The coefficient of {columns[i+1]} using the huber loss function is {deScaledCoefs[i]} {units[i+1]}')
+
+part1()
+part2()
+part3()
+part4()