Steamlit webapp for Steel plant load

Perdition

Inspiration

Which times of the year is the most energy consumed?

What patterns can we identify in energy usage?

About Dataset
Date Continuous-time data taken on the first of the month

Usage_kWh Industry Energy Consumption Continuous kWh

Lagging Current reactive power Continuous kVarh

Leading Current reactive power Continuous kVarh

Leading Current reactive power Continuous kVarh

CO2 Continuous ppm

NSM Number of Seconds from midnight Continuous S

Week status Categorical (Weekend (0) or a Weekday(1))

Day of week Categorical Sunday, Monday : Saturday

Load Type Categorical Light Load, Medium Load, Maximum Load

from IPython.display import YouTubeVideo

YouTubeVideo('3kXF5AMzcn0', width=950, height=400)

Streamlit website Link for Steel plant load Prediction

import pandas as pd
import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt
%matplotlib inline
from lightgbm import LGBMClassifier
from sklearn.linear_model import RidgeClassifierCV
from xgboost import XGBClassifier
from sklearn.neighbors import NearestCentroid
from sklearn.discriminant_analysis import QuadraticDiscriminantAnalysis
from sklearn.calibration import CalibratedClassifierCV
from sklearn.naive_bayes import BernoulliNB
from sklearn.ensemble import BaggingClassifier
from sklearn.linear_model import LogisticRegression
from sklearn.svm import SVC
from sklearn.svm import LinearSVC
from sklearn.neighbors import KNeighborsClassifier
from sklearn.naive_bayes import GaussianNB
from sklearn.linear_model import Perceptron
from sklearn.linear_model import SGDClassifier
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import RandomForestClassifier
from sklearn.neural_network import MLPClassifier
import xgboost as Xgb
from sklearn.ensemble import ExtraTreesClassifier
from sklearn.ensemble import AdaBoostClassifier
import lightgbm as lgb
from sklearn.svm import NuSVC
from sklearn.experimental import enable_hist_gradient_boosting
from sklearn.gaussian_process import GaussianProcessClassifier
from sklearn.gaussian_process.kernels import RBF
from sklearn.linear_model import RidgeClassifier
from sklearn.calibration import CalibratedClassifierCV
from sklearn.linear_model import PassiveAggressiveClassifier

from sklearn.metrics import accuracy_score, classification_report,f1_score,confusion_matrix,precision_score,recall_sc

from sklearn.metrics import confusion_matrix, plot_confusion_matrix, plot_roc_curve, plot_precision_recall_curve

import warnings
warnings.filterwarnings("ignore")

/opt/conda/lib/python3.7/site-packages/sklearn/experimental/enable_hist_gradient_boosting.py:17: UserWarning: S
ince version 1.0, it is not needed to import enable_hist_gradient_boosting anymore. HistGradientBoostingClassif
ier and HistGradientBoostingRegressor are now stable and can be normally imported from sklearn.ensemble.
"Since version 1.0, "

# Color Palettes
colors = ["#bfd3e6", "#9b5b4f", "#4e4151", "#dbba78", "#bb9c55", "#909195","#dc1e1e","#a02933","#716807","#717cb4"
sns.palplot(sns.color_palette(colors))

#Default theme
sns.set_theme(palette='tab10',
font='Comic Sans MS',
font_scale=1.5,
rc=None)

import matplotlib
matplotlib.rcParams.update({'font.size': 15})
plt.style.use('dark_background')
plt.rcParams["axes.grid"] = False

df = pd.read_csv("/kaggle/input/steel-industry-energy-consumption/Steel_industry_data.csv")
df.head().style.background_gradient(cmap='copper').set_precision(2)

date Usage_kWh Lagging_Current_Reactive.Power_kVarh Leading_Current_Reactive_Power_kVarh CO2(tCO2) Lagging_Current_Power_Facto

01/01/2018
0 3.17 2.95 0.00 0.00 73.21
00:15

01/01/2018
1 4.00 4.46 0.00 0.00 66.77
00:30

01/01/2018
2 3.24 3.28 0.00 0.00 70.28
00:45

01/01/2018
3 3.31 3.56 0.00 0.00 68.09
01:00

01/01/2018
4 3.82 4.50 0.00 0.00 64.72
01:15

df.describe().T.round(2).sort_values(by='std' , ascending = False)\

.style.background_gradient(cmap='GnBu')\
.bar(subset=["max"], color='#BB0000')\
.bar(subset=["min",], color='green')\
.bar(subset=["mean",], color='Orange')\
.bar(subset=['std'], color='#716807')\
 .bar(subset=['50%'], color='#717cb4')

count mean std min 25% 50% 75%

NSM 35040.000000 42750.000000 24940.530000 0.000000 21375.000000 42750.000000 64125.000000 85500.00000

Usage_kWh 35040.000000 27.390000 33.440000 0.000000 3.200000 4.570000 51.240000 157.18000

Leading_Current_Power_Factor 35040.000000 84.370000 30.460000 0.000000 99.700000 100.000000 100.000000 100.00000

Lagging_Current_Power_Factor 35040.000000 80.580000 18.920000 0.000000 63.320000 87.960000 99.020000 100.00000

Lagging_Current_Reactive.Power_kVarh 35040.000000 13.040000 16.310000 0.000000 2.300000 5.000000 22.640000 96.91000

Leading_Current_Reactive_Power_kVarh 35040.000000 3.870000 7.420000 0.000000 0.000000 0.000000 2.090000 27.76000

CO2(tCO2) 35040.000000 0.010000 0.020000 0.000000 0.000000 0.000000 0.020000 0.07000

corr = df.corr()
df.corr().style.background_gradient(cmap='copper').set_precision(2)

Usage_kWh Lagging_Current_Reactive.Power_kVarh Leading_Current_Reactive_Power_kVarh CO2(tCO2) Lag

Usage_kWh 1.00 0.90 -0.32 0.99

Lagging_Current_Reactive.Power_kVarh 0.90 1.00 -0.41 0.89

Leading_Current_Reactive_Power_kVarh -0.32 -0.41 1.00 -0.33

CO2(tCO2) 0.99 0.89 -0.33 1.00

Lagging_Current_Power_Factor 0.39 0.14 0.53 0.38

Leading_Current_Power_Factor 0.35 0.41 -0.94 0.36

NSM 0.23 0.08 0.37 0.23

df1 = df.copy()
#Correlation with Response Variable class
X = df1.drop(['Usage_kWh'],axis=1)
y = df1['Usage_kWh']

X.corrwith(y).plot.bar(
figsize = (16, 5), title = "Correlation with Steel plant Load Distribution", fontsize = 15,
rot = 45, grid = False)
plt.show()

class color:
BOLD = '\033[1m'
END = '\033[0m'
print(f"\033[94m\033[1m")
print(color.BOLD + 'Missing values - Percentage: \n' + color.END)
print(f"\033[91m\033[1m")
print(round(df.isnull().mean() * 100, 2))
Missing values - Percentage:

date 0.0
Usage_kWh 0.0
Lagging_Current_Reactive.Power_kVarh 0.0
Leading_Current_Reactive_Power_kVarh 0.0
CO2(tCO2) 0.0
Lagging_Current_Power_Factor 0.0
Leading_Current_Power_Factor 0.0
NSM 0.0
WeekStatus 0.0
Day_of_week 0.0
Load_Type 0.0
dtype: float64

cat = df.select_dtypes(include='object').columns.tolist()

for col in df[cat]:

print(f"\033[94m\033[1m")
print(col,"\n")
print(f"\033[91m\033[1m")
print(df[col].value_counts())
print(f"\033[92m\033[1m")
print("======="*5)

date

01/01/2018 00:15 1
01/09/2018 08:45 1
01/09/2018 07:15 1
01/09/2018 07:30 1
01/09/2018 07:45 1
..
02/05/2018 14:45 1
02/05/2018 14:30 1
02/05/2018 14:15 1
02/05/2018 14:00 1
31/12/2018 00:00 1
Name: date, Length: 35040, dtype: int64

===================================

WeekStatus

Weekday 25056
Weekend 9984
Name: WeekStatus, dtype: int64

===================================

Day_of_week

Monday 5088
Tuesday 4992
Wednesday 4992
Thursday 4992
Friday 4992
Saturday 4992
Sunday 4992
Name: Day_of_week, dtype: int64

===================================

Load_Type

Light_Load 18072
Medium_Load 9696
Maximum_Load 7272
Name: Load_Type, dtype: int64

===================================

#Rename some columns

df = df.rename(columns={'Lagging_Current_Reactive.Power_kVarh': 'Lagging_Reactive_Power_kVarh',
'Leading_Current_Reactive_Power_kVarh': 'Leading_Reactive_Power_kVarh',
'Lagging_Current_Power_Factor': 'Lagging_Power_Factor',
'Leading_Current_Power_Factor': 'Leading_Power_Factor',
'CO2(tCO2)':'CO2'})
df.head()
 date Usage_kWh Lagging_Reactive_Power_kVarh Leading_Reactive_Power_kVarh CO2 Lagging_Power_Factor Leading_Power_Factor NSM

01/01/2018
0 3.17 2.95 0.0 0.0 73.21 100.0
00:15

01/01/2018
1 4.00 4.46 0.0 0.0 66.77 100.0 1800
00:30

01/01/2018
2 3.24 3.28 0.0 0.0 70.28 100.0 2700
00:45

01/01/2018
3 3.31 3.56 0.0 0.0 68.09 100.0 3600
01:00

01/01/2018
4 3.82 4.50 0.0 0.0 64.72 100.0 4500
01:15

sns.displot(data=df, x="Lagging_Reactive_Power_kVarh", kde=True, bins = 100,color = "red", facecolor = "#3F7F7F"

plt.figure(figsize=(18,4))
color = plt.cm.copper(np.linspace(0, 1, 10))
df.groupby(['WeekStatus','Day_of_week'])['Usage_kWh'].count().plot(kind='bar', width=.4,color=color);
plt.xticks(rotation=45);

fig, (ax1,ax2) =plt.subplots(1,2, figsize=(18,4))

fig, (ax3,ax4) =plt.subplots(1,2, figsize=(18,4))
fig, (ax5,ax6) =plt.subplots(1,2, figsize=(18,4))
fig, (ax7,ax8) =plt.subplots(1,2, figsize=(18,4))

ax1.scatter(data=df,x="Usage_kWh", y="Lagging_Reactive_Power_kVarh", color=colors[1])

ax1.set_title("Usage kWh vs Lagging Reactive Power kVarh",pad=20)
ax1.set_xlabel("Usage (kWh)")
ax1.set_ylabel("Lagging Reactive Power (kVarh)")

ax2.scatter(data=df,x="Usage_kWh",y="Leading_Reactive_Power_kVarh", color=colors[8])
ax2.set_title("Usage(kWh) vs Leading Reactive Power(kVarh)",pad=20)
ax2.set_xlabel("Usage(kWh)")
ax2.set_ylabel("Leading Reactive Power (kVarh)")

ax3.scatter(data=df,x="Usage_kWh", y="Lagging_Power_Factor", color=colors[3])

ax3.set_title("Usage kWh vs Lagging Power Factor",pad=20)
ax3.set_xlabel("Usage (kWh)")
ax3.set_ylabel("Lagging Power Factor")

ax4.scatter(data=df,x="Usage_kWh",y="Leading_Power_Factor", color=colors[9])
ax4.set_title("Usage(kWh) vs Leading Power Factor",pad=20)
ax4.set_xlabel("Usage(kWh)")
ax4.set_ylabel("Leading Power Factor")
ax5.scatter(data=df,x="Lagging_Reactive_Power_kVarh",y="Leading_Reactive_Power_kVarh", color=colors[2])
ax5.set_title("Lagging Reactive Power (kVarh) vs Leading Reactive Power(kVarh)",pad=20,fontsize=15)
ax5.set_xlabel("Lagging Reactive Power (kVarh)")
ax5.set_ylabel("Leading Reactive Power(kVarh)")

ax6.scatter(data=df,x="Lagging_Power_Factor",y="Leading_Power_Factor", color=colors[4])
ax6.set_title("Lagging Power Factor vs Leading Power Factor",pad=20,fontsize=15)
ax6.set_xlabel("Lagging Power Factor")
ax6.set_ylabel("Leading Power Factor")

ax7.scatter(data=df,x="Lagging_Reactive_Power_kVarh",y="Lagging_Power_Factor", color=colors[5])
ax7.set_title("Lagging Reactive Power (kVarh) vs Leading Power Factor",pad=20,fontsize=15)
ax7.set_xlabel("Lagging Reactive Power (kVarh)")
ax7.set_ylabel("Leading Power Factor")

ax8.scatter(data=df,x="Lagging_Reactive_Power_kVarh",y="Leading_Power_Factor", color=colors[4])
ax8.set_title("Lagging Reactive Power (kVarh) vs Leading Power Factor",pad=20,fontsize=15)
ax8.set_xlabel("Lagging Reactive Power (kVarh)")
ax8.set_ylabel("Leading Power Factor")

plt.show()

sns.relplot(data=df, x="Usage_kWh", y="Lagging_Reactive_Power_kVarh", hue="WeekStatus",col="WeekStatus",palette=

sns.relplot(data=df, x="Usage_kWh", y="Leading_Reactive_Power_kVarh", hue="WeekStatus",col="WeekStatus",palette=
sns.relplot(data=df, x="Usage_kWh", y="Lagging_Power_Factor", hue="WeekStatus",col="WeekStatus",palette='tab10');
sns.relplot(data=df, x="Usage_kWh", y="Leading_Power_Factor", hue="WeekStatus",col="WeekStatus",palette='tab10');
sns.relplot(data=df, x="Usage_kWh", y="Lagging_Reactive_Power_kVarh", hue="Load_Type",col="Load_Type",palette='tab10'
sns.relplot(data=df, x="Usage_kWh", y="Leading_Reactive_Power_kVarh", hue="Load_Type",col="Load_Type",palette='tab10'
sns.relplot(data=df, x="Usage_kWh", y="Lagging_Power_Factor", hue="Load_Type",col="Load_Type",palette='tab10');
sns.relplot(data=df, x="Usage_kWh", y="Leading_Power_Factor", hue="Load_Type",col="Load_Type",palette='tab10');
ax = plt.figure(figsize=(18,6))
ax = plt.subplot(1,2,1)
ax = sns.countplot(x='Load_Type', data=df)
ax.bar_label(ax.containers[0])
plt.title("Load_Type", fontsize=20,color='#dbba78',font='Times New Roman',pad=30)
ax =plt.subplot(1,2,2)
ax=df['Load_Type'].value_counts().plot.pie(explode=[0.1, 0.1,0.1],autopct='%1.2f%%',shadow=True);
ax.set_title(label = "Load_Type", fontsize = 20,color='#dbba78',font='Times New Roman',pad=30);

from statsmodels.graphics.gofplots import qqplot

var = df['Lagging_Reactive_Power_kVarh']
color = colors[4]
fig = plt.figure(figsize = (18, 12))

# --- Title ---

plt.title("Lagging Reactive Power kVarh Distribution",fontsize=20,font='Comic Sans MS',pad=40,color = colors[1])

# --- Histogram ---

ax_1=fig.add_subplot(2, 2, 2)
plt.title('Histogram Plot', fontweight = 'bold', fontsize = 14, fontfamily = 'Comic Sans MS', color = colors[1])
sns.histplot(data = df, x = var, kde = True, color = color)
plt.xlabel('Total', fontweight = 'regular', fontsize = 11, fontfamily = 'Comic Sans MS', color = colors[1])
plt.ylabel('Lagging_Reactive_Power_kVarh', fontweight = 'regular', fontsize = 11, fontfamily = 'sans-serif', color
plt.grid(axis = 'x', alpha = 0)
plt.grid(axis = 'y', alpha = 0.2)

# --- Q-Q Plot ---

ax_2 = fig.add_subplot(2, 2, 4)
plt.title('Q-Q Plot', fontweight = 'bold', fontsize = 14, fontfamily = 'Comic Sans MS', color = colors[1])
qqplot(var, fit = True, line = '45', ax = ax_2, markerfacecolor = color, markeredgecolor = color, alpha = 0.6)
plt.xlabel('Theoritical Quantiles', fontweight = 'regular', fontsize = 11, fontfamily = 'Comic Sans MS',
color = colors[3])
plt.ylabel('Sample Quantiles', fontweight = 'regular', fontsize = 11, fontfamily = 'Comic Sans MS', color = colors

# --- Boxen Plot ---

ax_3 = fig.add_subplot(1, 2, 1)
plt.title('Boxen Plot', fontweight = 'bold', fontsize = 14, fontfamily = 'Comic Sans MS', color = colors[3])
sns.boxenplot(y = var, data = df, color = color, linewidth = 1.5)
plt.ylabel('Lagging_Reactive_Power_kVarh', fontweight = 'regular', fontsize = 11, fontfamily = 'Comic Sans MS',

plt.show();

plt.figure(figsize=(18,7))
sns.scatterplot(data=df, x="Usage_kWh", y="Lagging_Reactive_Power_kVarh", hue="Load_Type",palette="tab10");

plt.figure(figsize=(18,10))

plt.subplot(2,2,1)
sns.barplot(x = 'Load_Type', y = 'Usage_kWh', palette= "tab10",data=df)
plt.title("Load Type", color = "#bfd3e6")
plt.xlabel("Load_Type")
plt.ylabel("Usage_kWh")

plt.subplot(2,2,2)
df["WeekStatus"].value_counts().plot.pie(autopct='%1.2f%%', explode=[0.1, 0.1], colors=['blue','#dbba78'])
p = plt.gcf()
plt.title("Weekday/Weekend")
plt.legend()
plt.subplot(2,2,(3,4))

sns.countplot(x = 'WeekStatus', hue = 'Load_Type', data = df, palette="twilight")

plt.title("Usage kWh by Load Type", color = "Lightpink")
plt.xlabel("Load_Type")
plt.tight_layout()
plt.show()

col = ['Usage_kWh']
fig, ax = plt.subplots(2, 1, sharex=True, figsize=(17,8),gridspec_kw={"height_ratios": (.2, .8)})
ax[0].set_title('Usage_kWh distribution',fontsize=18,pad=20)
sns.boxplot(x='Usage_kWh', data=df, ax=ax[0])
ax[0].set(yticks=[])
sns.histplot(x='Usage_kWh', data=df, ax=ax[1])
ax[1].set_xlabel(col, fontsize=16)
plt.axvline(df['Usage_kWh'].mean(), color='darkgreen', linewidth=2.2, label='mean=' + str(np.round(df['Usage_kWh'
plt.axvline(df['Usage_kWh'].median(), color='red', linewidth=2.2, label='median='+ str(np.round(df['Usage_kWh'].
plt.axvline(df['Usage_kWh'].mode()[0], color='purple', linewidth=2.2, label='mode='+ str(df['Usage_kWh'].mode()[
plt.legend(bbox_to_anchor=(1, 1.03), ncol=1, fontsize=17, fancybox=True, shadow=True, frameon=True)
plt.tight_layout()
plt.show()

#Disaster count(Whether the disaster happened or not )

#Disaster count(Whether the disaster happened or not )
plt.figure(figsize = (18, 8))
ax = plt.axes()
ax.set_facecolor('black')
ax = sns.countplot(x = 'Load_Type', data = df, palette = 'tab10', edgecolor = 'white', linewidth = 1.2)
plt.title('Load Types', fontsize = 25)
plt.xlabel('Load Type', fontsize = 20)
plt.ylabel('Count', fontsize = 20)
ax.xaxis.set_tick_params(labelsize = 15)
ax.yaxis.set_tick_params(labelsize = 15)
bbox_args = dict(boxstyle = 'round', fc = '0.9')
for p in ax.patches:
ax.annotate('{:.0f} = {:.2f}%'.format(p.get_height(), (p.get_height() / len(df['Load_Type'])) * 100), (p
color = 'black',
bbox = bbox_args,
fontsize = 13)
plt.show()

col_names = ['Lagging_Reactive_Power_kVarh','Leading_Reactive_Power_kVarh','Lagging_Power_Factor','Leading_Power_Fact
fig, axs = plt.subplots(nrows=2,ncols=3,figsize=(20,10))
for i in range(0, len(col_names)):
rows = i // 3
cols = i % 3
ax = axs[rows,cols]
plot = sns.regplot(x = col_names[i], y = 'Usage_kWh', data = df, ax = ax )

var = ['Usage_kWh', 'Lagging_Reactive_Power_kVarh','Leading_Reactive_Power_kVarh','Lagging_Power_Factor','Leading_Pow

from scipy.stats import skew

for col in df[var]:

print(f"\033[91m\033[1m")
print("Skewness:",col,"=",round(skew(df[col]),3))
print("Kurtosis:",col, "=",round(df[col].kurt(),2))
 print("Mean:",col, "=",round(df[col].mean(),2))
print("Max:",col, "=",round(df[col].max(),2))
print("Min:",col, "=",round(df[col].min(),2))
print("Median:",col, "=",round(df[col].median(),2))
print("Std:",col, "=",round(df[col].std(),2))
print("Var:",col, "=",round(df[col].var(),2))
print("Mode:",col, "=",round(df[col].mode(),2))
plt.figure(figsize=(18,6))
sns.distplot(df[col],kde=True,bins=50,color="Yellow",hist_kws={"edgecolor": (1,1,0,1)})
plt.title(col,fontweight="bold")
plt.show()
print(f"\033[93m\033[1m")
print("====="*25)

Skewness: Usage_kWh = 1.197

Kurtosis: Usage_kWh = 0.39
Mean: Usage_kWh = 27.39
Max: Usage_kWh = 157.18
Min: Usage_kWh = 0.0
Median: Usage_kWh = 4.57
Std: Usage_kWh = 33.44
Var: Usage_kWh = 1118.53
Mode: Usage_kWh = 0 3.06
dtype: float64

===============================================================================================================
==============

Skewness: Lagging_Reactive_Power_kVarh = 1.438

Kurtosis: Lagging_Reactive_Power_kVarh = 1.21
Mean: Lagging_Reactive_Power_kVarh = 13.04
Max: Lagging_Reactive_Power_kVarh = 96.91
Min: Lagging_Reactive_Power_kVarh = 0.0
Median: Lagging_Reactive_Power_kVarh = 5.0
Std: Lagging_Reactive_Power_kVarh = 16.31
Var: Lagging_Reactive_Power_kVarh = 265.89
Mode: Lagging_Reactive_Power_kVarh = 0 0.0
dtype: float64
===============================================================================================================
==============

Skewness: Leading_Reactive_Power_kVarh = 1.734

Kurtosis: Leading_Reactive_Power_kVarh = 1.58
Mean: Leading_Reactive_Power_kVarh = 3.87
Max: Leading_Reactive_Power_kVarh = 27.76

Min: Leading_Reactive_Power_kVarh = 0.0

Median: Leading_Reactive_Power_kVarh = 0.0
Std: Leading_Reactive_Power_kVarh = 7.42
Var: Leading_Reactive_Power_kVarh = 55.12
Mode: Leading_Reactive_Power_kVarh = 0 0.0
dtype: float64

===============================================================================================================
==============

Skewness: Lagging_Power_Factor = -0.606

Kurtosis: Lagging_Power_Factor = -1.1
Mean: Lagging_Power_Factor = 80.58
Max: Lagging_Power_Factor = 100.0
Min: Lagging_Power_Factor = 0.0
Median: Lagging_Power_Factor = 87.96
Std: Lagging_Power_Factor = 18.92
Var: Lagging_Power_Factor = 358.02
Mode: Lagging_Power_Factor = 0 100.0
dtype: float64

===============================================================================================================
==============

Skewness: Leading_Power_Factor = -1.512

Kurtosis: Leading_Power_Factor = 0.38
Mean: Leading_Power_Factor = 84.37
Max: Leading_Power_Factor = 100.0
Min: Leading_Power_Factor = 0.0
Median: Leading_Power_Factor = 100.0
Std: Leading_Power_Factor = 30.46
Var: Leading_Power_Factor = 927.6
Mode: Leading_Power_Factor = 0 100.0
dtype: float64
===============================================================================================================
==============

from sklearn.preprocessing import FunctionTransformer

from sklearn.compose import ColumnTransformer

old_skew = df.skew().sort_values(ascending=False)
old_skew

def logTrans(feature): # function to apply transformer and check the distribution with histogram and kdeplot

logTr = ColumnTransformer(transformers=[("lg", FunctionTransformer(np.log1p), [feature])])

plt.figure(figsize=(15,6))
plt.subplot(1,2,1)
plt.title("Distribution before Transformation", fontsize=20,color='red')
sns.histplot(df[feature], kde=True, color="red")
plt.xlabel(feature,color='Red')

plt.subplot(1,2,2)
df_log = pd.DataFrame(logTr.fit_transform(df))
plt.title("Distribution after Transformation", fontsize=20,color='Blue')
sns.histplot(df_log,bins=20, kde=True , legend=False)
plt.xlabel(feature,color='Blue')
plt.show()

print(f"Skewness was {round(old_skew[feature],2)} before & is {round(df_log.skew()[0],2)} after Log transformatio

logTrans(feature="Lagging_Reactive_Power_kVarh")

Skewness was 1.44 before & is -0.02 after Log transformation.

plt.figure(figsize=(18,4))
sns.kdeplot(data=df,x="Usage_kWh",hue='Load_Type',multiple="stack");
# Encode Categorical Columns
from sklearn.preprocessing import LabelEncoder
categ = df.select_dtypes(include = "object").columns

le = LabelEncoder()
df[categ] = df[categ].apply(le.fit_transform)

df.head()

date Usage_kWh Lagging_Reactive_Power_kVarh Leading_Reactive_Power_kVarh CO2 Lagging_Power_Factor Leading_Power_Factor NSM

0 1 3.17 2.95 0.0 0.0 73.21 100.0 900

1 2 4.00 4.46 0.0 0.0 66.77 100.0 1800

2 3 3.24 3.28 0.0 0.0 70.28 100.0 2700

3 4 3.31 3.56 0.0 0.0 68.09 100.0 3600

4 5 3.82 4.50 0.0 0.0 64.72 100.0 4500

# Split the dataset and prepare some lists to store the models
from sklearn.model_selection import train_test_split
X = df.drop(['Load_Type'], axis=1)
y = df.Load_Type

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.25, random_state = 42)

models = []

names = [
"LGBMClassifier",
"RidgeClassifierCV",
"XGBClassifier",
"QuadraticDiscriminantAnalysis",
"CalibratedClassifierCV",
"BernoulliNB",
"BaggingClassifier",
"LogisticRegression",
"NearestCentroid",
"SVC",
"LinearSVC",
"KNeighborsClassifier",
"GaussianNB",
"Perceptron",
"SGDClassifier",
"DecisionTreeClassifier",
"RandomForestClassifier",
"MLPClassifier",
"ExtraTreesClassifier",
"AdaBoostClassifier",
"NuSVC"
]

scores = []

clf =[
LGBMClassifier(),
RidgeClassifierCV(),
XGBClassifier(),
QuadraticDiscriminantAnalysis(),
CalibratedClassifierCV(),
BernoulliNB(),
BaggingClassifier(),
LogisticRegression(),
NearestCentroid(),
SVC(),
LinearSVC(),
KNeighborsClassifier(),
GaussianNB(),
Perceptron(),
 SGDClassifier(),
DecisionTreeClassifier(),
RandomForestClassifier(),
MLPClassifier(),
ExtraTreesClassifier(),
AdaBoostClassifier(),
NuSVC()

%%time
for model in clf:
model.fit(X_train, y_train)
score = model.score(X_test, y_test)
scores.append(score)

final_scores = pd.DataFrame(zip(names,scores), columns=['Classifier', 'Accuracy'])

CPU times: user 2min 57s, sys: 2.79 s, total: 3min

Wall time: 2min 53s

final_scores.sort_values(by='Accuracy',ascending=False).style.background_gradient(cmap="copper").set_properties(
'font-family': 'Comic Sans MS',
'color': 'Brown',
'font-size': '15px',"color": "Brown"
})

Classifier Accuracy

2 XGBClassifier 0.965068

0 LGBMClassifier 0.956963

6 BaggingClassifier 0.952740

15 DecisionTreeClassifier 0.936872

16 RandomForestClassifier 0.935274

18 ExtraTreesClassifier 0.921918

19 AdaBoostClassifier 0.855594

11 KNeighborsClassifier 0.796119

1 RidgeClassifierCV 0.744064

9 SVC 0.736986

20 NuSVC 0.729566

3 QuadraticDiscriminantAnalysis 0.713128

5 BernoulliNB 0.704452

12 GaussianNB 0.691210

7 LogisticRegression 0.671461

8 NearestCentroid 0.665525

10 LinearSVC 0.632534

17 MLPClassifier 0.613356

4 CalibratedClassifierCV 0.606050

14 SGDClassifier 0.605137

13 Perceptron 0.524087

p = plt.figure(figsize=(18,20))
p = sns.set_context('paper', font_scale=1.8)

p = final_scores=final_scores.sort_values(by='Accuracy',ascending=False)[:20]

p = sns.barplot(y= 'Classifier', x= 'Accuracy', data= final_scores,palette=colors[0:10])

for container in p.containers:

p.bar_label(container,label_type = 'center',padding = 15,size = 20,color = "Red",rotation = 0,
bbox={"boxstyle": "round", "pad": 0.6, "facecolor": "#a9a9a9", "edgecolor": "white", "alpha": .8})
plt.title('COMPARE THE MODEL',fontsize=20,color='#013220')
plt.xlabel('MODEL',fontsize=20,color='#013220')
plt.ylabel('Model Accuracy',fontsize=20,color='#013220');

Streamlit website Link for Steel plant load

Prediction
from IPython.display import YouTubeVideo
YouTubeVideo('3kXF5AMzcn0', width=950, height=400)
Loading [MathJax]/extensions/Safe.js