About
Articles by Vishal
Activity
10K followers
Experience & Education
Licenses & Certifications
Volunteer Experience
-
Volunteer
Boys and Girls Clubs of America
- Present 11 years 8 months
Children
-
Volunteer
Red cross
Health
Projects
-
Topic Modeling for Customer Complaints
-
See projectThis project involves using LDA (Latent Dirichilet Allocation) algorithm to uncover hidden topics from customer complaints data. The complaints data was obtained from the Consumer Finance Protection Bureau website (https://www.consumerfinance.gov/data-research/consumer-complaints/) with a focus on credit card products.
Results: The model was used to categorize customer complaints into more accurate and intuitive categories, reducing rerouting of complaints and uncovering hidden themes in…This project involves using LDA (Latent Dirichilet Allocation) algorithm to uncover hidden topics from customer complaints data. The complaints data was obtained from the Consumer Finance Protection Bureau website (https://www.consumerfinance.gov/data-research/consumer-complaints/) with a focus on credit card products.
Results: The model was used to categorize customer complaints into more accurate and intuitive categories, reducing rerouting of complaints and uncovering hidden themes in the data.
Python Libraries used: Numpy, Pandas, NLTK, SpaCy, Gensim, Mallet, pyLDAvis -
Machine Learning for Fraud Detection with SMOTE, AutoEncoders and Isolation Forest
-
See projectObjective: To build a predictive model for detecting credit card fraud using supervised and unsupervised Machine Learning and Deep Learning algorithms.
Data: Public dataset from Kaggle Challenge (https://www.kaggle.com/mlg-ulb/creditcardfraud). It has 284K samples and 30 independent variables.
Algorithms: Supervised ML - Random Forest, XGBoost, LightGBM with SMOTE (Synthetic Minority Oversampling TEchniques) for handling class imbalance Deep Learning-based Methods - AutoEncoders…Objective: To build a predictive model for detecting credit card fraud using supervised and unsupervised Machine Learning and Deep Learning algorithms.
Data: Public dataset from Kaggle Challenge (https://www.kaggle.com/mlg-ulb/creditcardfraud). It has 284K samples and 30 independent variables.
Algorithms: Supervised ML - Random Forest, XGBoost, LightGBM with SMOTE (Synthetic Minority Oversampling TEchniques) for handling class imbalance Deep Learning-based Methods - AutoEncoders Unsupervised ML - Local Outlier Factor, Isolation Forest
Conclusion: Fraud detection is one of the most interesting and complex Machine Learning problems. Typically it involves extreme class imbalance which needs to be handled by techniques such as SMOTE. Also, credit card fraud is an ever-evolving beast like Hydra! If you cut one head, 2 other heads will grow back in its place! This means a supervised learning framework does not work well because the model performance deteriorates very quickly. Unsupervised anomaly detection algorithms such as Isolation Forest show a great potential to develop a more robust fraud detection system
Honors & Awards
-
2019 Ones to Watch Award
CIO & the CIO Executive Council
About the Ones to Watch Awards:
The Ones to Watch awards spotlight rising technology leaders who have what it takes to become strategic, C-level business technology executives of tomorrow. Candidates have a track record of success, an essential blend of communication and collaboration skills and a keen understanding of business goals.
About CIO:
CIO focuses on attracting the highest concentration of enterprise CIOs and business technology executives with unparalleled peer insight…About the Ones to Watch Awards:
The Ones to Watch awards spotlight rising technology leaders who have what it takes to become strategic, C-level business technology executives of tomorrow. Candidates have a track record of success, an essential blend of communication and collaboration skills and a keen understanding of business goals.
About CIO:
CIO focuses on attracting the highest concentration of enterprise CIOs and business technology executives with unparalleled peer insight and expertise on business strategy, innovation, and leadership. As organizations grow with digital transformation, CIO provides its readers with key insights on career development, including certifications, hiring practices and skills development. The award-winning CIO portfolio provides business technology leaders with analysis and insight on information technology trends and a keen understanding of IT’s role in achieving business goals. More information on CIO is available at www.cio.com. -
Loyalty360 Platinum Award for Loyalty Analytics
Loyalty360
Languages
-
English
Full professional proficiency
-
Hindi
Full professional proficiency
-
Marathi
Full professional proficiency
Recommendations received
10 people have recommended Vishal
Join now to viewOther similar profiles
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content