NYC Data Science Academy| Blog
Bootcamps
Lifetime Job Support Available Financing Available
Bootcamps
Data Science with Machine Learning Flagship ๐Ÿ† Data Analytics Bootcamp Artificial Intelligence Bootcamp New Release ๐ŸŽ‰
Free Lesson
Intro to Data Science New Release ๐ŸŽ‰
Find Inspiration
Find Alumni with Similar Background
Job Outlook
Occupational Outlook Graduate Outcomes Must See ๐Ÿ”ฅ
Alumni
Success Stories Testimonials Alumni Directory Alumni Exclusive Study Program
Courses
View Bundled Courses
Financing Available
Bootcamp Prep Popular ๐Ÿ”ฅ Data Science Mastery Data Science Launchpad with Python View AI Courses Generative AI for Everyone New ๐ŸŽ‰ Generative AI for Finance New ๐ŸŽ‰ Generative AI for Marketing New ๐ŸŽ‰
Bundle Up
Learn More and Save More
Combination of data science courses.
View Data Science Courses
Beginner
Introductory Python
Intermediate
Data Science Python: Data Analysis and Visualization Popular ๐Ÿ”ฅ Data Science R: Data Analysis and Visualization
Advanced
Data Science Python: Machine Learning Popular ๐Ÿ”ฅ Data Science R: Machine Learning Designing and Implementing Production MLOps New ๐ŸŽ‰ Natural Language Processing for Production (NLP) New ๐ŸŽ‰
Find Inspiration
Get Course Recommendation Must Try ๐Ÿ’Ž An Ultimate Guide to Become a Data Scientist
For Companies
For Companies
Corporate Offerings Hiring Partners Candidate Portfolio Hire Our Graduates
Students Work
Students Work
All Posts Capstone Data Visualization Machine Learning Python Projects R Projects
Tutorials
About
About
About Us Accreditation Contact Us Join Us FAQ Webinars Subscription An Ultimate Guide to
Become a Data Scientist
    Login
NYC Data Science Acedemy
Bootcamps
Courses
Students Work
About
Bootcamps
Bootcamps
Data Science with Machine Learning Flagship
Data Analytics Bootcamp
Artificial Intelligence Bootcamp New Release ๐ŸŽ‰
Free Lessons
Intro to Data Science New Release ๐ŸŽ‰
Find Inspiration
Find Alumni with Similar Background
Job Outlook
Occupational Outlook
Graduate Outcomes Must See ๐Ÿ”ฅ
Alumni
Success Stories
Testimonials
Alumni Directory
Alumni Exclusive Study Program
Courses
Bundles
financing available
View All Bundles
Bootcamp Prep
Data Science Mastery
Data Science Launchpad with Python NEW!
View AI Courses
Generative AI for Everyone
Generative AI for Finance
Generative AI for Marketing
View Data Science Courses
View All Professional Development Courses
Beginner
Introductory Python
Intermediate
Python: Data Analysis and Visualization
R: Data Analysis and Visualization
Advanced
Python: Machine Learning
R: Machine Learning
Designing and Implementing Production MLOps
Natural Language Processing for Production (NLP)
For Companies
Corporate Offerings
Hiring Partners
Candidate Portfolio
Hire Our Graduates
Students Work
All Posts
Capstone
Data Visualization
Machine Learning
Python Projects
R Projects
About
Accreditation
About Us
Contact Us
Join Us
FAQ
Webinars
Subscription
An Ultimate Guide to Become a Data Scientist
Tutorials
Data Analytics
  • Learn Pandas
  • Learn NumPy
  • Learn SciPy
  • Learn Matplotlib
Machine Learning
  • Boosting
  • Random Forest
  • Linear Regression
  • Decision Tree
  • PCA
Interview by Companies
  • JPMC
  • Google
  • Facebook
Artificial Intelligence
  • Learn Generative AI
  • Learn ChatGPT-3.5
  • Learn ChatGPT-4
  • Learn Google Bard
Coding
  • Learn Python
  • Learn SQL
  • Learn MySQL
  • Learn NoSQL
  • Learn PySpark
  • Learn PyTorch
Interview Questions
  • Python Hard
  • R Easy
  • R Hard
  • SQL Easy
  • SQL Hard
  • Python Easy
Data Science Blog > Machine Learning > Higgs Boson Machine Learning Challenge

Higgs Boson Machine Learning Challenge

Belinda Kanpetch, Taraqur Rahman, Steven Ginzberg and Michael Winfield
Posted on Jun 10, 2016

Contributed by by Belinda Kanpetch, Taraqur Rahman, Steven Ginzberg and Michael Winfield.  They are currently in the NYC Data Science Academy 12 week full time Data Science Bootcamp program taking place between January 11th to April 1st, 2016. This post is based on their fourth class project - Machine learning(due on the 8th week of the program).

Introduction

The Higgs Boson Challenge, hosted by Kaggle, asked the data scientist community to utilize machine learning to accurately predict if a particle was a Higgs-Boson particle or not; more specifically if a signal detected was either a โ€˜tau tau decay of a Higgs bosonโ€™ or just โ€˜backgroundโ€™.

The datasets provided were the training and test set with 250,000 and 550,000 observations, respectively. The training set contained all the same features as the test with two additional columns of โ€˜Labelโ€™ and โ€˜Weightโ€™ that gave the accurate classifiers to help train our models.

Our approach for this project was to get a little background on the physics and cursory understanding of how the experiment was performed.  We looked at correlations between the variables and noticed that a few were highly correlated and thus began researching how other scientists had suggested working with the correlated variables.  This lead us to the feature engineering discussed below. In addition, we all utilized the Caret package in R for all coding.

Check out our github to see the code.

The sections below will unpack:

  • How we approached missing values
  • Feature engineering
  • XGBoost model
  • Gradient Boosting model
  • Random Forests
  • Neural Networks
  • Conclusion & Next steps

Missingness & Feature Engineering

The missing values in the dataset (-999) resulted from two sources:

  1. Bad estimates of the mass of Higgs boson; and
  2. Jets: particles that can appear 0, 1, 2, or 3 times in an event.

To deal with this structural missingness, we:

  • Converted the -999s into NAs.
  • Created a Boolean column to reflect missing(T)/present(F) values in the estimated mass of the Higgs boson (DER_mass_MMC).
  • Created 8 binary columns reflecting each combination of presence(0)/absence(1) in the estimated mass of the Higgs boson and the number of jets (0,1,2,3).
  • o J0 +M1, J0+M0, J1+M1, J1+M0โ€ฆ
  • Imputed the column mean for all the NAs.

For more on the structural nature of the missing values in the Higgs boson dataset, see cowa14.pdf.

Feature Selection

To deal with the risk of overfitting, we:

  • Subtracted the PRI_tau_phi column from the other 4 โ€˜phiโ€™ columns.
  • Rotated the angle of the remaining 4 phi columns.
  • Deleted all 5 of the non-rotated phi columns, leaving us with 4 phi columns instead of 5.

This subtraction + rotation process makes the specific pattern of the phi variables more unique, so that, once the model is trained, it is better able to discriminate between signal and background in the test set.

For more on the logic of this process, see diaz14.pdf

QJ_PRESENT_Page_05 QJ_PRESENT_Page_06 QJ_PRESENT_Page_07

XGBoost

Our team explored XGBoost for a few reasons. First, it ran relatively quickly; and second there seemed to be a fair number of tuning parameters available to play with.
Playing with the model and actually getting it to a good fit are two completely different concepts.  In general, we split the training set: 80% of the training set to train the model, and the remaining 20% as a test set.  In all cases we used the AMS as the deciding metric, and measured accuracy as the percentage of the 20% test data correctly IDโ€™d as our  performance indicator.
Some of the tuning parameters used were:
โ€ข    Number of cross-validation folds, between 2 and 5
โ€ข    Number of rounds, between 50 and 400
โ€ข    Eta (shrinkage) between .001 and .6
โ€ข    Gamma between .01 and .6
โ€ข    Max tree depth between 6 and 10
Generally, we received AMS scores on the training (80%) set of between 1.2 all the way up to 1.8+.  These resulted in accuracy rates of 80%+ when applied to the test (20%) dataset.  Our first instinct is this was overfitting, and that turned out to be the case.  The AMS scores on the Kaggle full test data site were much lower, with the highest being about .5.

QJ_PRESENT_Page_10 QJ_PRESENT_Page_09

Gradient Boosting Model

We implemented a GBM to compare with the other non-linear models performance.  Our intuition was that this would be a good comparison with the Random Forest since they are both tree based methods but are grown in different ways; with GBM taking into account the mistakes or residuals of the previous trees. We had to be cognizant of overfitting and adjusting parameters to make the learn the model slowly.

The train.csv file we received from Kaggle was split into an 80% train and 20% test sets. Using the Caret package we implemented an expanded grid of parameters to train the model.

This returned an optimal model with parameters and score below.

QJ_PRESENT_Page_13

Although the accuracy was better than a 50/50 chance it did not perform as well as we had anticipated. Therefore, we trained another model with a larger number of trees and larger lambda, or shrinkage, to try have the tree learn slower.

Of the parameters we specified, an optimal model with parameters and score were returned below.

QJ_PRESENT_Page_14

It is interesting to note that even though the second model returned a lower accuracy rate the Kaggle score was higher most likely due to overfitting.

Some challenges were the lack of computational power as we very soon learned that running GBM was computational expensive and took hours to train one model.  Additionally, in hindsight we should have started running a substantially larger number of trees.

Random Forest

For our RF model, we tried both bagging (mtry=38) and a number of other mtry values (7, 12, 38).  We trained on the entire training dataset and used the out of bag error instead of cross-validating with 80% of the dataset.

Overfitting due to leakage was not an issue with the model due to the feature engineering on the dataset that substantially reduced its multicollinearity.

Using parallel processing and a server enabled this model to run well in under 3 hours.

In hindsight, we could have used the number of trees for the RF model as a benchmark for determining the point at which a boosted trees model outperforms the RF, rather than allowing the Caret package to select the optimal number of trees itself.

Neural Network

We applied the neural networks to our dataset that we feature engineered. We believed the weight decay was crucial for the neural network. The weight decay is a multiplier that determines the significance of each node in the network. If it is extremely small then that means that node has no effect on the output. If it were big then the node would have a significant impact on the output. During the process of neural networks, each input goes through a node. When there are multiple nodes, one way the network knows which node to take precedent is by assigning a โ€œweightโ€ or value to it depending on how significant that node is for the prediction. The weight decay is a useful multiplier that neural networks use to automatically go backwards to adjust the weights if they are not that significant in coming up with a prediction.

For the first neural network trial, we tried to implement three hidden layers of twenty nodes each. Our understanding was that since weight decay is applied, we would not have to worry about overfitting. However we were wrong. This model produced a threshold of 0.002 (graph of Specificity vs. Sensitivity). Specificity is the false negatives while the Sensitivity is the true positives. The maximum point where the sensitivity and specificity were optimal was at 0.002 (threshold).

QJ_PRESENT_Page_21

Learning from our mistake, we chose different hidden layers since even with the weight decay, overfitting is possible. This time we used one hidden layer with 20 nodes. This produced a better threshold of 0.041 and a better accuracy.

QJ_PRESENT_Page_22

During our research we came across another neural technique called dropout neural network. Based on our understanding, in an ideal scenario, neural networks would like to go through each node and see what its exact weight would be and average all of them. However it is computationally expensive. The way the dropout method works is that it randomly โ€œdropsโ€ a node and then calculate the weight of the others. Once it gets to the end, it will go backwards and drop another node and adjust the weights accordingly. With this technique, there are fewer nodes to calculate and therefore it would be close to the ideal scenario without the expensive computations. This works because all the nodes have a fifty percent chance of dropping which prevents the technique from being bias. We believed this would be great to calculate for the AMS.

Conclusions and Next Steps

What we would like to do next:

  • Look into ensembling models
  • Exploration of other packages other than Caret.
  • With more time, we could have written a function that:
    • Identifies values of PRI_tau_eta < 0
    • Converts the signs of all eta values in the same row (- to +, + to -)
    • Imputes the column mean for missing values and randomly assigns a sign (+,-)
  • http://www.jmlr.org/proceedings/papers/v42/diaz14.pdf

Model Comparison Sheet

QJ_model comparison sheet

About Authors

Belinda Kanpetch

Belinda hails from the Bayou City, where she earned her Bachelors of Architecture from the University of Houston. Drawn to the bright lights and energy of the city, she relocated to New York to attend Columbia University at...
View all posts by Belinda Kanpetch >

Taraqur Rahman

During his career as a Sales Associate, Taraqur analyzed data to help support both the sales and marketing teams. Seeing through his own eyes how much data can influence decisions, Taraqur joined NYCDSA as a data scientist in...
View all posts by Taraqur Rahman >

Steven Ginzberg

Steven has spent a number of years performing systems development, financial analysis and management in a variety of companies. Most recently, Steven has been working with start-ups helping them go from conception of ideas, identifying technologies, and finally...
View all posts by Steven Ginzberg >

Michael Winfield

Michael has a passion for finding strategic insights for businesses, managers, and organizations engaged in competitive dynamics. With a background in corporate litigation and white collar criminal defense, as well as graduate-level education in strategic management, Michael is...
View all posts by Michael Winfield >

Related Articles

Capstone
Catching Fraud in the Healthcare System
Data Analysis
Car Sales Report R Shiny App
Data Analysis
Injury Analysis of Soccer Players with Python
Capstone
Acquisition Due Dilligence Automation for Smaller Firms
R Shiny
Forecasting NY State Tax Credits: R Shiny App for Businesses

Leave a Comment

Cancel reply

You must be logged in to post a comment.

No comments found.

View Posts by Categories

All Posts 2399 posts
AI 7 posts
AI Agent 2 posts
AI-based hotel recommendation 1 posts
AIForGood 1 posts
Alumni 60 posts
Animated Maps 1 posts
APIs 41 posts
Artificial Intelligence 2 posts
Artificial Intelligence 2 posts
AWS 13 posts
Banking 1 posts
Big Data 50 posts
Branch Analysis 1 posts
Capstone 206 posts
Career Education 7 posts
CLIP 1 posts
Community 72 posts
Congestion Zone 1 posts
Content Recommendation 1 posts
Cosine SImilarity 1 posts
Data Analysis 5 posts
Data Engineering 1 posts
Data Engineering 3 posts
Data Science 7 posts
Data Science News and Sharing 73 posts
Data Visualization 324 posts
Events 5 posts
Featured 37 posts
Function calling 1 posts
FutureTech 1 posts
Generative AI 5 posts
Hadoop 13 posts
Image Classification 1 posts
Innovation 2 posts
Kmeans Cluster 1 posts
LLM 6 posts
Machine Learning 364 posts
Marketing 1 posts
Meetup 144 posts
MLOPs 1 posts
Model Deployment 1 posts
Nagamas69 1 posts
NLP 1 posts
OpenAI 5 posts
OpenNYC Data 1 posts
pySpark 1 posts
Python 16 posts
Python 458 posts
Python data analysis 4 posts
Python Shiny 2 posts
R 404 posts
R Data Analysis 1 posts
R Shiny 560 posts
R Visualization 445 posts
RAG 1 posts
RoBERTa 1 posts
semantic rearch 2 posts
Spark 17 posts
SQL 1 posts
Streamlit 2 posts
Student Works 1687 posts
Tableau 12 posts
TensorFlow 3 posts
Traffic 1 posts
User Preference Modeling 1 posts
Vector database 2 posts
Web Scraping 483 posts
wukong138 1 posts

Our Recent Popular Posts

AI 4 AI: ChatGPT Unifies My Blog Posts
by Vinod Chugani
Dec 18, 2022
Meet Your Machine Learning Mentors: Kyle Gallatin
by Vivian Zhang
Nov 4, 2020
NICU Admissions and CCHD: Predicting Based on Data Analysis
by Paul Lee, Aron Berke, Bee Kim, Bettina Meier and Ira Villar
Jan 7, 2020

View Posts by Tags

#python #trainwithnycdsa 2019 2020 Revenue 3-points agriculture air quality airbnb airline alcohol Alex Baransky algorithm alumni Alumni Interview Alumni Reviews Alumni Spotlight alumni story Alumnus ames dataset ames housing dataset apartment rent API Application artist aws bank loans beautiful soup Best Bootcamp Best Data Science 2019 Best Data Science Bootcamp Best Data Science Bootcamp 2020 Best Ranked Big Data Book Launch Book-Signing bootcamp Bootcamp Alumni Bootcamp Prep boston safety Bundles cake recipe California Cancer Research capstone car price Career Career Day ChatGPT citibike classic cars classpass clustering Coding Course Demo Course Report covid 19 credit credit card crime frequency crops D3.js data data analysis Data Analyst data analytics data for tripadvisor reviews data science Data Science Academy Data Science Bootcamp Data science jobs Data Science Reviews Data Scientist Data Scientist Jobs data visualization database Deep Learning Demo Day Discount disney dplyr drug data e-commerce economy employee employee burnout employer networking environment feature engineering Finance Financial Data Science fitness studio Flask flight delay football gbm Get Hired ggplot2 googleVis H20 Hadoop hallmark holiday movie happiness healthcare frauds higgs boson Hiring hiring partner events Hiring Partners hotels housing housing data housing predictions housing price hy-vee Income industry Industry Experts Injuries Instructor Blog Instructor Interview insurance italki Job Job Placement Jobs Jon Krohn JP Morgan Chase Kaggle Kickstarter las vegas airport lasso regression Lead Data Scienctist Lead Data Scientist leaflet league linear regression Logistic Regression machine learning Maps market matplotlib Medical Research Meet the team meetup methal health miami beach movie music Napoli NBA netflix Networking neural network Neural networks New Courses NHL nlp NYC NYC Data Science nyc data science academy NYC Open Data nyc property NYCDSA NYCDSA Alumni Online Online Bootcamp Online Training Open Data painter pandas Part-time performance phoenix pollutants Portfolio Development precision measurement prediction Prework Programming public safety PwC python Python Data Analysis python machine learning python scrapy python web scraping python webscraping Python Workshop R R Data Analysis R language R Programming R Shiny r studio R Visualization R Workshop R-bloggers random forest Ranking recommendation recommendation system regression Remote remote data science bootcamp Scrapy scrapy visualization seaborn seafood type Selenium sentiment analysis sentiment classification Shiny Shiny Dashboard Spark Special Special Summer Sports statistics streaming Student Interview Student Showcase SVM Switchup Tableau teachers team team performance TensorFlow Testimonial tf-idf Top Data Science Bootcamp Top manufacturing companies Transfers tweets twitter videos visualization wallstreet wallstreetbets web scraping Weekend Course What to expect whiskey whiskeyadvocate wildfire word cloud word2vec XGBoost yelp youtube trending ZORI

NYC Data Science Academy

NYC Data Science Academy teaches data science, trains companies and their employees to better profit from data, excels at big data project consulting, and connects trained Data Scientists to our industry.

NYC Data Science Academy is licensed by New York State Education Department.

Get detailed curriculum information about our
amazing bootcamp!

Please enter a valid email address
Sign up completed. Thank you!

Offerings

  • HOME
  • DATA SCIENCE BOOTCAMP
  • ONLINE DATA SCIENCE BOOTCAMP
  • Professional Development Courses
  • CORPORATE OFFERINGS
  • HIRING PARTNERS
  • About

  • About Us
  • Alumni
  • Blog
  • FAQ
  • Contact Us
  • Refund Policy
  • Join Us
  • SOCIAL MEDIA

    ยฉ 2025 NYC Data Science Academy
    All rights reserved. | Site Map
    Privacy Policy | Terms of Service
    Bootcamp Application