Predicting

IPL
Auction Prices

Introduction to Business Analytics

By: Arun 22037
Ayush Kumar 22044
Abhishek Modi 22010
Vibhor Kumar 22270
INTRODUCTION

Overview:
The primary aim of this project revolves around leveraging machine learning techniques to
predict auction prices of IPL (Indian Premier League) players based on their attributes. By
employing regression analysis on player data, we aim to estimate the expected price at which
a player might be auctioned.

Importance:
• Strategic Financial Decision-making: Predicting auction prices holds immense value for
IPL teams in formulating their financial strategies. It assists franchises in budget allocation
and optimal player selection during auctions.
• Player Valuation: Understanding player values through statistical modeling aids teams in
assessing player worth, negotiating contracts, and forming a balanced team composition
for the tournament.

Objectives:
• Auction Price Forecasting: Utilize attributes like nationality, player type, batting average,
and bowling average to predict the probable auction price of IPL players.
• Regression Techniques Demonstration: Showcase the practical application of regression
algorithms in forecasting player values, demonstrating the predictive capabilities of
machine learning in a real-world sports context.
DATASET

Our Dataset contains the 7 columns and 205 rows. For determining the Price of the
Players in IPL Auction we need to consider some of the important factors like Nationality,
Type, Status(Capped/Uncapped), Batting Average, Bowling Average. These factors are
given larger importance while determining the price of the Players in IPL Auction.

Factors Details
Nationality The player's country of origin or nationality (e.g., Indian, Overseas).
Players Type The player's role or category in cricket (e.g., Batsman, Bowler, All-
Rounder).
Status Indicates whether the player is categorized as Capped (having
International Level experience) or Uncapped .
Batting Average The average number of runs scored by the player per dismissal while
batting in T20s.
Bowling Average The average number of runs conceded by the player per wicket taken
while bowling in T20s.
METHODOLOGY

• Importing Data: Utilized Pandas library to load IPL auction dataset ('IPL.xlsx') into a Pandas DataFrame for further analysis.
Data • Exploratory Data Analysis (EDA):
• Inspected data structure using df.info() to comprehend column data types, missing values, and dataset size.
Loadi • Employed visualizations like scatter plots and pair plots using Seaborn and Matplotlib libraries to identify relationships and
distributions among attributes.
ng & • Created a heatmap to visually represent correlations among attributes using sns.heatmap(df.iloc[:,[5,6,7]].corr(), annot=True).
Prepr
ocessi
ng
• Data Preparation:
• Segregated dataset into features (independent variables) and the target variable (Sold_Price).
• Split the dataset into training and testing sets using train_test_split from sklearn.model_selection.
• Feature Encoding:
Model •Utilized OneHotEncoder and ColumnTransformer to encode categorical features (Nationality, Type, Status) into numerical format
suitable for model training.
Devel • Model Building:

opme •Employed a Linear Regression model via a pipeline using make_pipeline and LinearRegression.
•Fitted the model with training data using pipe.fit(X_train, Y_train).
nt

•Prediction for New Player Data:

•Created a new dataset for a hypothetical player and used the model to predict the auction price for this
player.
Predic •Model Performance Evaluation:
tion & •Assessed the model's performance using evaluation metrics such as Mean Squared Error (MSE) and R-
squared (mean_squared_error, r2_score).
Evalua
tion
FINDINGS
Pair Plot of Dataset
FINDINGS

Correlation Matrix Heatmap Actual vs Predicted Prices on Test Set

LIMITATIONS AND SCOPE FOR IMPROVEMNENT
Enhancement Opportunities
• Feature Expansion: Consider augmenting the
Feature Set Limitation: The model's predictive feature set by incorporating additional relevant
power might be constrained by the limited set of attributes, such as player performance in specific
features considered (Nationality, Type, Batting match formats, recent form, injury history, or
Average, Bowling Average). The exclusion of other team dynamics, to enhance predictive accuracy.
potential influential factors could impact its • Improved Data Collection: Focus on acquiring
accuracy in estimating auction prices. more comprehensive and diverse data sources to
mitigate missing data issues and provide a more
holistic view of player attributes and performance
metrics.
• Advanced Modeling Techniques: Explore the
application of more sophisticated modeling
Data Quality: Potential limitations in data quality, approaches beyond linear regression, such as
such as missing or incomplete information, might ensemble methods (Random Forests, Gradient
affect the model's performance and predictive Boosting), which could potentially capture
nonlinear relationships more effectively.
capability. These data inadequacies might lead to
biased or less accurate predictions.

Validation and Iteration

• Model Validation: Implement rigorous validation
techniques, such as cross-validation, to assess
model stability and generalizability, ensuring that
Model Complexity: The model's simplicity (linear the model performs consistently across various
regression) might limit its ability to capture datasets and scenarios.
intricate nonlinear relationships between • Iterative Approach: Adopt an iterative approach
attributes and auction prices. More complex to model development by continuously refining
the model based on new data insights and
models might better represent the underlying
feedback from predictions to enhance its accuracy
complexities. and reliability.
THANK
YOU!!
"In the ever-evolving world of cricket auctions, predictive models pave
the way for strategic decisions. Anticipating player values is not just a
statistic, but a game-changer in shaping winning teams."