Description

Sehban Fazili
M Tech CSE: MT21143
ML-based prediction system to predict the job role for a new graduate.
Steps:
1. Importing required Libraries from sklearn import metrics from sklearn.neural_network import MLPClassifier from sklearn.model_selection import train_test_split import matplotlib.pyplot as plt import seaborn as sns import numpy as np import pandas as pd from sklearn.preprocessing import LabelEncoder, OneHotEncoder from sklearn.preprocessing import Normalizer
2. Reading the dataset dataset = pd.read_csv(“roo_data.csv”)
3. Exploratory Data Analysis
In the given dataset:
Total entries = 20000 from range 0 to 19999 Null Values = None
dtypes: int64(14), object(25)i.e.,
Categorical Data count: 25
Numerical Data count: 14
3.1 Dropping irrelevant columns using drop commandfinal = dataset.drop(drop_columns, axis = 1)
3.2 Since the first 9 columns are Percentages of a student in different subjects so we take the mean of that and print a single percentage column with the name ‘Academic Percentage’.
In this step, first percentage columns were only taken using the ‘iloc’ function call, then mean was calculated using df.mean(axis=1)command.
Finally from the created data frame, only the column with mean values was kept and the rest were dropped. The data frame was named df1.
3.3 A separate data frame was created ( not with percentages) df3.
Then df1 and df3 were concatenated using the command final_df = pd.concat([df1,df3],axis=1)
3.4 Reduced the classes to a total of 6 classes based on some similarities between them.
3.5 Encoding was done on categorical data points using LabelEncoder &
OneHotEncoder.
labelencoder = LabelEncoder() for i in range(4,9):
data[:,i] = labelencoder.fit_transform(data[:,i])
3.6 Normalization was done to scale down the data using the commandnormalized_data = Normalizer().fit_transform(data1) Now dataset can be worked on.
4. Classification Task
4.1 Splitting the dataset into train and test using the command:
X_train,X_test,y_train,y_test=train_test_split(X1,y,test_size=
0.2,stratify=y,random_state=10)
4.2 Applying Multi-layer Perceptron classifier on the train and test data and fitting the model using the command:
classifier = MLPClassifier(hidden_layer_sizes=(40,10,2), max_iter=300,activation = ‘relu’,solver=’adam’,random_state=1) classifier.fit(X_train, y_train)
4.3 Predicting the model and calculating accuracy, confusion matrix, etc.
5. Results:
Train-Test Splits Accuracy
80-20 31.275
70-30 31.2
60-40 31.1875
90-10 31.25
Classification Report:
1) 80-20 split

2) 70-30 split

3) 60-40 split

4) 90-10 split

6. Code Listing:
from sklearn import metrics from sklearn.neural_network import MLPClassifier from sklearn.model_selection import train_test_split import matplotlib.pyplot as plt import seaborn as sns import numpy as np import pandas as pd
dataset = pd.read_csv(“roo_data.csv”)
dataset.info()
dataset.describe()
drop_columns = [‘Hours working per day’,’Management or Technical’,’hackathons’,’can work long time before system?’,’Extra-courses did’,’workshops’,’worked in teams ever?’,’self-learning capability?’,’memory capability score’,’Job/Higher Studies?’,’hard/smart worker’,’talenttests taken?’,’olympiads’,’reading and writing skills’,’Type of company want to settle in?’,’interested in games’,’Taken inputs from seniors or elders’,’Interested Type of Books’,’In a
Realtionship?’,’Gentle or Tuff behaviour?’,’Introvert’] final = dataset.drop(drop_columns, axis = 1)
final.head()
df = final.iloc[:,:9]
df.head()
df[‘Average Academic Percentage’] = df.mean(axis=1)
df.head()
drop_list = [‘Acedamic percentage in Operating Systems’, ‘percentage in Algorithms’
,’Percentage in Programming Concepts’,’Percentage in Software Engineering’, ‘Percentage in
Computer Networks’, ‘Percentage in Electronics Subjects’ ,’Percentage in Computer
Architecture’, ‘Percentage in Mathematics’, ‘Percentage in Communication skills’]
df1 = df.drop(drop_list, axis = 1)
df1.info()
df1.head()
df2 = df1[‘Average Academic Percentage’].nunique() df2
df3 = final.iloc[:,9:]
df3.head()
final_df = pd.concat([df1,df3],axis=1)
final_df.head()
final_df.shape
Suggested_DF = final_df.iloc[:,9:]
Suggested_DF.head()
Suggested_DF[‘Suggested Job Role’].replace({“Database Developer”: ‘Database
Profile’,’Database Manager’:’Database Profile’,’Database Administrator’:’Database Profile’,’Data
Architect’: “Database Profile”, ‘Technical Support’:’Technical/Support Profile’,’Technical Services/Help Desk/Tech Support’:’Technical/Support Profile’,’Information Technology
Auditor’:’Technical/Support Profile’,’Software Quality Assurance (QA) /
Testing’:’Technical/Support Profile’,’Information Technology Manager’:’Technical/Support
Profile’,’Technical Engineer’:’Technical/Support Profile’,’Portal Administrator’:’Technical/Support
Profile’,’Quality Assurance Associate’:’Technical/Support Profile’,’Systems
Analyst’:’Technical/Support Profile’,’Solutions Architect’:’Technical/Support Profile’,’Systems Security Administrator’:’Networking Profile’,’Network Security Administrator’:’Networking Profile’,’Network Engineer’:’Networking Profile’,’Network Security Engineer’:’Networking
Profile’,’Information Security Analyst’:’Networking Profile’,’UX Designer’:’Design Profile’,’Design
& UX’:’Design Profile’,”Business Systems Analyst”: “Business Profile”, “Business Intelligence Analyst”: “Business Profile”, “Project Manager”:”Business Profile”,”E-Commerce Analyst”: “Business Profile”,”CRM Technical Developer”: ‘Technical/Support Profile’, “CRM Business Analyst”: “Business Profile”,”Software Systems Engineer”: “Software Profile”,’Programmer Analyst’:’Software Profile’,”Mobile Applications Developer”: “Software Profile”, “Web Developer”:
“Software Profile”, “Software Developer”: “Software Profile”, “Applications Developer”: “Software
Profile”, “Software Engineer”: “Software Profile”}, inplace=True)
Suggested_DF.head()
Suggested_DF[‘Suggested Job Role’].nunique()
first_half=final_df.iloc[:,:9]
first_half.head()
roo_datset = pd.concat([first_half,Suggested_DF], axis=1)
roo_datset.shape
data = roo_datset.iloc[:,:-1].values label = roo_datset.iloc[:,-1].values len(data[0])
roo_datset.iloc[:,4:9]
roo_datset.iloc[:,:4]
from sklearn.preprocessing import LabelEncoder, OneHotEncoder
labelencoder = LabelEncoder()
for i in range(4,9):
data[:,i] = labelencoder.fit_transform(data[:,i])
data[:5]
data[:5,4:]
from sklearn.preprocessing import Normalizer
data1=data[:,:4]
normalized_data = Normalizer().fit_transform(data1) print(normalized_data.shape)
normalized_data
data2=data[:,4:] data2.shape
df11 = np.append(normalized_data,data2,axis=1)
df11.shape
X1 = pd.DataFrame(df11,columns=[‘Average Academic Percentage’,’Logical quotient
rating’,’coding skills rating’,’public speaking points’,’certifications’,’Interested subjects’, ‘interested career area’,’Salary Range Expected’, ‘Salary/work’]) X1.head()
label = labelencoder.fit_transform(label) print(len(label))
y=pd.DataFrame(label,columns=[“Suggested Job Role”]) y.head()
from sklearn import tree
from sklearn.model_selection import train_test_split from sklearn import preprocessing from sklearn.metrics import accuracy_score from sklearn.metrics import confusion_matrix
X_train,X_test,y_train,y_test=train_test_split(X1,y,test_size=0.2,stratify=y,random_state=10) “””MLP Classifier”””
#Importing MLPClassifier from sklearn.neural_network import MLPClassifier
#Initializing the MLPClassifier
classifier = MLPClassifier(hidden_layer_sizes=(40,10,2), max_iter=300,activation =
‘relu’,solver=’adam’,random_state=1)
classifier.fit(X_train, y_train)
#Predicting y for X_val y_pred = classifier.predict(X_test)
cm = confusion_matrix(y_test,y_pred) accuracy = accuracy_score(y_test,y_pred)
print(“confusion matrics=”,cm) print(” “)
print(“accuracy=”,accuracy*100)
from sklearn.metrics import classification_report
print(classification_report(y_test, y_pred))