Mohanasundaram Sithanathan

Project Name Victor Predictor:

It is an updated version of Ask Kail project but with so many changes we've added a new features it will be biased point of view predictions of the team which user has the authority to choose the team after choosing the team we'll be providing only about that team's perspective to win the match. We've purchased the Ronauz api for providing the details about the matches, teams and players. Also we've added the betting into it were we'll be predicting the bets we can use it in Bet Way website and make money to our users.

In this project my works are:

1.Pre-Match Predictions

In Previous version we've predicted the Pre-Match Insights based only on the past data but here we've added the Match Simulation method were we will simulate the match as the actual cricket match like a virtual match based on past data in ball by ball.

1. Top batting shots of the batsman and top bowling styles of the bowler.

2. Top Partnership Combinations in team1 and Weak Bowlers to target from team2.

3. Player vs player will give the best suitable bowler to get the batsman out or to make pressure on him while he is on strike this considered both wicket percentage and dot percentage.

4. Under Ratted players including new players from both the teams.

5. Target winning score prediction for 1st batting team.

2.Live Predictions

We've converted everything into visual representation compared to the previous version we'll be providing it at the end of every single over. Also we've added an engaging factor for users an Anticipation Intelligence we'll be predicting what will happen in the coming overs.

1. Players Performance BPM (Benchmark Performance Metric)

2. Run Rate Graph

3. in-game Winning Probability

4. Next Over Predictions (4s, 6s, maidens and wickets)

Project Name Ask Kail:

This is a Chabot cricket website the main functionalities of this project is we are providing the Chabot with the Pre-Match, In-Game and Post-Match insights and analysis using Flask and HTML we’ve integrated this and Deployment of it.

In this project my works are:

1.In Pre-Match Insights and Predictions were we've to provide this before the match day to users.

1. Predicted Winning Probability for both teams

2. Target Runs Prediction

3. Target Defending

4. Top Partnership Combinations

5. Best bowler to attack the batsman

2.In-Game Insights and Predictions were we've to provide this while the game is going on live

1. Ball by ball live commentary from cricbuzz website

2. Batting style and Bowling style predictions based on the commentary

3. Providing top best shots for the specific bowling style to batsman

4. Next over predictions

3.Post-Game Insights were we'll provide after the couple of matches played:

1. Series Winner or Tournament Winner

Project Name Marketing Mix Model(MMM):

First time I got an opportunity to work with real time data in this project. The real time data in raw with most data attributes were missing and not properly listed. I've used imputations and done all the pre-processing methods and made it reliable to perform analysis. My Objective for this project is to answer the below three questions.

1. What are the key drivers that affect sales?

I've found the significance of data based on hypothesis testing by using the p-value method.

2. What is the sales decomposition % of the key drivers?

This was found out by the regression coefficients of their each attributes that are significance with the sales.

3. How much money spent on the key drivers to find out how much revenue they have got?

I've used different type of modeling’s to predict the revenue of the sales are since we got multiple independent variables and dependent variable with continuous values we are using these models. Multiple Regression model (Ordinary Least Square), Support Vector Machine (Linear, Polynomial and RBF) and Ensemble Models (Bagging(Random Forest) and Boosting(Extreme Gradient Boosting)) were I've concluded that Extreme Gradient Boost algorithm has the lesser error rate and had high model performance compared to the other models that I’ve used above.