My Portfolio

Hi! My name is Chanakarn Luangpirom and this is my portfolio. This page demonstrate my past
data science related projects that I've worked on. It contains a brief summary of what I've done and also
a link to each individual project repository for more details and code!

July 2021

Trading Bot Deployed on AWS

  • Trading bot deployed on AWS utilizing ECR (Elastic Container Registry) to deploy a docker image.
  • Trade Strategy: RSI (Relative Strength Index) overbought/oversold.
  • Utilize AWS EventBridge to execute trade strategy when market open.
  • Use Amazon SNS to Send SMS to phone tracking daily portfolio change.
  • Store data using Amazon RDS logging trade positions into the database for further analytics.
  • Connect Analytics tool like tableau for analysis.
August 2020

Salary Prediction Chatbot

  • Scraped around 10,000 job postings from jobsdb using python and BeautifulSoup. :Data Scraping Colab
  • Perform multiple text data cleaning such as salary & working experience extraction from text using regex.
  • Perform exploratory data analysis to gain insights for jobs using plotly & seaborn. : EDA & Cleaning Colab
  • Create a tree-based model to predict job salary given experience year & job function.
  • Built two salary prediction chatbot & models. (1.Using local database 2.Using Firebase)
  • Built a flask API endpoint hosted locally and use ngrok to communicate with dialogflow to identify user’s intent.
April 2020

Solar Irradiance Forecasting

  • Using sensor measurement data collected from year 2017-2018 to predict solar irradiance
  • Perform data imputation based on analysis and assumptions e.g. Linear Interpolation, Column Deletions,
    Row Deletions, Replace with Average.
  • Perform exploratory data analysis identifying seasonality and irradiance at different time of day using python matplotlib.
  • Using a unique train-validation-test spliting method to adjust for data's seasonality nature.
  • Compare 3 types prediction methods (Baseline, Linear, Tree-based) and identify their pros and cons based on mean bias error (MBE) analysis.
June - August 2020

Shopee Data Analytics
Code League
(Sentiment Analysis)

  • 2nd place out of 356 teams (top 1%)
  • Given product reviews, build a model to correctly predict user ratings (1 to 5) for the review.
  • Competition Length : 2 weeks
  • Perfom multiple text data cleaning, undersamling, and an ensemble of GPT-2, Roberta, and BERT models to predict ratings.
June - August 2020

Shopee Data Analytics
Code League
(Marketing Analytics)

  • 16th place out of 368 teams (top 7%)
  • Build a model that can predict whether a user opens the emails sent by Shopee.
  • Competition Length : 1 week
  • Perform tabular data cleaning and applied Tomeklinks undersampling method to help model generalize & optimize probability cut-off threshold on catboost model to obtain the best Matthews correlation coefficient.
June - August 2020

Shopee Data Analytics
Code League
(Product Detection)

  • 63rd out of 823 teams (top 8%)
  • Build a multiple image classification model. Dataset include ~100k images within 42 different categories.
  • Competition Length : 2 weeks
  • Perform image augmentation and ensemble of resnet, inception, inceptionresnet models for final predictions.