You are on page 1of 14

PROJECT WORK ON

‘COUNTER STRIKE: GO – ROUND


WINNER CLASSIFICATION’
USING ANN IN AZURE ML
Background: CS: GO is a shooting game, where 2 teams Terrorist and CT play
for a best of 30 rounds, with every round being 1:55 minute. There are a total
of 5 players on each team (10 in total) and the 1st team to reach 16 rounds
wins the game. At the start, one team plays as Terrorist and the other as CT.
After 15 rounds played, the squads swap side. There are 7 unique maps a game
can be played on. You conquest a round as Terrorist by either planting the
explosives and making sure it explodes, or by eliminating the other squad. You
win a round as CT by either eliminating the other squad, or by disarming the
explosive, should it have been planted.
Purpose: This project work is intended to classify the CS: GO round winner as
per player’s details.
Case Study: The case study is about classification of the CS: GO round winner
using ANN (Artificial Neural Networks) in Microsoft Azure ML Studio. Kaggle
link for the project is attached below.
https://www.kaggle.com/christianlillelund/csgo-round-winner-classification

Note:- Dataset used in this project is provided along with the ques document.

You have to create an Artificial Neural networks classification model on Azure


ML and publish the web-service. Your output should be the classification of CS:
GO round winner.

Steps Solutions:
Ques1: How to upload the datasets to Azure ML?
Solution1:
Upload the data files you have downloaded in .csv format. Go to DATASETS tab
given on left corner of your screen Click on “+ NEW” button Select “From a
Local File” option and then upload the data file from your system to AZURE
server. [Follow the Fig.1 and Fig.2 below.]
Fig.1

Fig.2

Ques2: How to create new experiment?


Solution2:
Create a new experiment (By clicking on “+ NEW” tab and choose Blank
Experiment Option) to build Prediction model. Drag your data files into the
newly created Experiment. Now, summarize your Dataset using ‘Summarize
Data’ from search bar  Drag ‘Summarize Data’ below your dataset  Select
the end node of dataset and connect it to the starting node of ‘Summarize
Data’ Click ‘RUN’. [Follow Fig.3 below.]

Fig.3

Now, right click and visualize the ‘Summarize Data’ to get 5-point summary of
the dataset i.e., Mean, Mode, Median, 1st Quartile, 3rd Quartile and the
missing, unique value count of dataset. [Follow the Fig.4, Fig.5 below.]
Fig.4

Fig.5

Ques3: How to Normalize the dataset?

Solution3:
Normalizing the Dataset
Search for ‘Normalize Data’  Drag and drop ‘Normalize Data’ below the
dataset and connect the nodes  Select the Transformation method from the
drop-down menu  Click on ‘Launch column selector’ and select all the
columns you want to normalize except the target column. [Follow the Fig.6,
Fig.7 below.]  Finally click ‘RUN’.
Fig.6

Fig.7

Ques4: How to split the dataset?

Solution4:
Split the Dataset
Search for ‘Split Data’  Drag and drop ‘Split Data’ below the ‘Normalize Data’
and connect the nodes  Type in the Fraction of rows section to split data in
desired ratio (in my case it’s 70:30). [Follow Fig.8 below.]  Finally click ‘RUN’.
Fig.8

Ques5: How to choose a modelling technique and apply to your model?

Solution5:
Choosing the modelling technique
Search for ‘Two-Class Neural Network’  Drag and drop ‘Two-Class Neural
Network’ beside the ‘Split Data’ Select the parameters i.e., Number of
hidden nodes, learning rate, Number of learning iterations, The initial learning
weight  Select ‘Do not normalize’ from the drop-down menu in ‘The type of
normalizer’. [Follow the Fig.9 below.]  Finally click ‘RUN’.

Hint: In my case these were the parameters: -


1. Number of hidden nodes – 200
2. Learning rate – 0.5
3. Number of learning iterations – 300
4. The initial learning weight – 0.3
Fig.9

Ques6: How to train the model?

Solution6:
Train the model
Search for ‘Train model’  Drag and drop ‘Train model’ below the ‘Split Data’
and ‘Two-Class Neural Network  Click on ‘Launch column selector’ Select
the target column and move it to right side. [Follow the Fig.10, Fig.11 below.]
 Click on the Tick  Finally click ‘RUN’.

Fig.10
Fig.11

Ques7: How to score the model?

Solution7:
Scoring the model
Search for ‘Score model’  Drag and drop ‘Score model’ twice (one for
scoring the train dataset and other for scoring the test dataset) below the
‘Train model’ Connect the 1st node of the ‘Score model’ with ‘Train model’
and 2nd node of ‘Score model’ with 1st node of ‘Split data’ for train scoring and
2nd node of ‘Score model’ with 2nd node of ‘Split data’ for test scoring [Follow
Fig.12, Fig.13 below.]Finally click ‘RUN’.
Fig.12

Ques8: How to evaluate the model?

Solution8:
Evaluating the model
Search for ‘Evaluate model’  Drag and drop ‘Evaluate model’ twice (one for
evaluating the train dataset and other for evaluating the test dataset) below
the ‘Score model’ Connect the node of the ‘Score model’ with ‘Evaluate
model’ for both train and test dataset [Follow Fig.13, Fig.14, Fig.15
below.]Finally click ‘RUN’ Right click on ‘Evaluate model’ and visualize to
separately to check the evaluation metric (In this project we are depending on
the Accuracy).
Fig.13

Fig.14
Fig.15

Ques9: How to deploy the web-service and test your model?

Solution9:
Deploy the Web-service and test your ANN model.
Click on ‘Setup web service’ Select ‘Create predictive model’  New tab will
be created i.e., ‘Predictive experiment’  Click on ‘Predictive experiment’ and
‘RUN’ the model
After Running the Predictive Experiment Deploy the Experiment (Click on the
“Deploy Web service” tab). [Follow the Fig.16, Fig.17 below.]
Fig.16

Fig.17

Test your recommendation engine by clicking on the link or click on the “Test”
button. Give the inputs and get the outputs. [Follow the Fig 18, Fig 19 below.]
Fig.18

Fig.19

----------------------End of Assignment----------------------

You might also like