Introduction

The project investigates the factors influencing which countries import the most commodities. With globalization, understanding export-import dynamics is crucial for optimizing trade strategies and supporting economic policies. Analyzing import patterns can provide insights into market dependencies, identify potential new markets, and enhance decision-making for exporters.

Dataset Description

Data collect data using the U.S. Census Bureau’s International Trade API, which provides detailed information on exports and imports, including commodity codes and values. This will serve as our primary data source.

• Dataset Source: U.S. Census Bureau’s International Trade API

• Dataset URL: “https://api.census.gov/data/timeseries/intltrade/exports/hs”

Variables Included

Total 41 Variables

• Export District Code: Unique identifier for the export district.
• District Name: Name of the export district (e.g., Baltimore, MD).
• Export Commodity Code: Code representing the specific commodity being exported.
• Export Commodity Long Description: Detailed description of the commodity.
• Export Commodity Short Description: Brief description of the commodity.
• Monthly Export Value (Total): Total value of exports for the month.
• Total Monthly Export Value: Cumulative total export value for the month.
• Year-to-Date Export Value (Total): Cumulative value of exports for the year up to the specified month.
• Total Year-to-Date Export Value: Overall cumulative export value for the year.
• Monthly Vessel Export Value: Value of exports transported by vessel for the month.
• Year-to-Date Vessel Export Value: Cumulative value of vessel exports for the year.
• Monthly Vessel Weight: Weight of exports transported by vessel for the month.
• Year-to-Date Vessel Weight: Cumulative weight of vessel exports for the year.
• Monthly Air Export Value: Value of exports transported by air for the month.
• Year-to-Date Air Export Value: Cumulative value of air exports for the year.
• Monthly Air Weight: Weight of exports transported by air for the month.
• Year-to-Date Air Weight: Cumulative weight of air exports for the year.
• Monthly Card Count: Count of export transactions for the month.
• Year-to-Date Card Count: Count of export transactions for the year.
• Monthly Quantity Value: Value of exported quantities for the month.
• Year-to-Date Quantity Value: Cumulative value of exported quantities for the year.
• Monthly Quantity Weight: Weight of exported quantities for the month.
• Year-to-Date Quantity Weight: Cumulative weight of exported quantities for the year.
• Commodity Level: Level of detail for the commodity classification.
• Country Code: Code representing the country of destination.
• Country Name: Name of the destination country.
• Domestic/Foreign Code: Indicator of whether the data is for domestic or foreign entities.
• Last Update Date: Date when the data was last updated.
• Month: The month of observation.
• Monthly Quantity 1: Quantity metric for the month.
• Quantity 1 Flag: Flag indicating the status of Quantity 1.
• Year-to-Date Quantity 1: Cumulative quantity metric for the year.
• Quantity 1 Year Flag: Flag indicating the status of Year-to-Date Quantity 1.
• Monthly Quantity 2: Second quantity metric for the month.
• Quantity 2 Flag: Flag indicating the status of Quantity 2.
• Year-to-Date Quantity 2: Cumulative second quantity metric for the year.
• Quantity 2 Year Flag: Flag indicating the status of Year-to-Date Quantity 2.
• Summary Level: Level of summarization for the data.
• Summary Level 2: Additional level of summarization.
• Unit Quantity 1: Measurement unit for Quantity 1.
• Unit Quantity 2: Measurement unit for Quantity 2.

Project Investigation

• In this project, we aim to analyze export data from the Baltimore, MD district to uncover insights related to export patterns, trends, and factors influencing export values. Using a dataset that includes various metrics, such as monthly and year-to-date export values, shipping methods, and commodity descriptions.

• This dataset enables to predict export values, for different commodities, transport modes across different countries.

Why is this data interesting?

• Dataset holds both time-series and categorical features, provide an opportunity to explore patterns over time and the potential effects of commodity type, country of origin, and mode of transport on export volumes.

• Export data holds both short-term and long-term patterns that vary by mode, which RNNs are well-suited to capture due to their ability to maintain and use memory over sequences.

• Performance of different modes under varying conditions is essential for logistics optimization helps forecast inventory needs, plan shipments and reduce supply chain bottlenecks.

Data Analysis Framework

Data Exploration(EDA):

• Summary Statistics: Generate descriptive statistics to capture the essence of the data.

• Visualizations: Create plots to visualize trends over time and identify any seasonal patterns in exports.

Data Wrangling:

• Addressing any missing or inconsistent values

• Converting categorical fields to factors for analytical use

• Creating derived metrics where applicable to enhance analysis

• Feature Engineering: Implement advanced feature engineering techniques to enhance model performance by creating new variables based on existing data.

Data Modelling:

• Utilize machine learning frameworks, specifically H2O and Keras, to predict future export values based on historical data.

• Deep Learning Models: Leverage neural networks tailored for structured data analysis, particularly focusing on the relationships among various export-related variables.

• Applying deep learning models like RNNs, CNNs, and bidirectional LSTMs enables the captures complex, nonlinear dependencies in processing sequential data.

• Dynamic Time Warping (DTW) techniques to measure similarity between time-series sequences to compare the temporal structure of sequences. Explore ways to combine Dynamic Time Warping (DTW) techniques and deep learning technique, to enhance model performance.

Model Evaluation:

Model’s performance is evaluated on the test set, using metrics such as RMSE, Adj-R2, MSE for different model and compare their performance.

if (!require('tidyverse')) install.packages('tidyverse'); library('tidyverse')
if (!require('tidyr')) install.packages('tidyr'); library('tidyr')
if (!require('kableExtra')) install.packages('kableExtra'); library('kableExtra')
if (!require('RColorBrewer')) install.packages('RColorBrewer'); library('RColorBrewer')
if (!require('ggplot2')) install.packages('ggplot2'); library('ggplot2')
if (!require('ggthemes')) install.packages('ggthemes'); library('ggthemes')
if (!require('httr')) install.packages('httr'); library('httr')
if (!require('jsonlite')) install.packages('jsonlite'); library('jsonlite')
if (!require('stringr')) install.packages('stringr'); library('stringr')
if (!require('skimr')) install.packages('skimr'); library('skimr')
if (!require('recipes')) install.packages('recipes'); library('recipes')
if (!require('tibble')) install.packages('tibble'); library('tibble')
if (!require('h2o')) install.packages('h2o'); library('h2o')
if (!require('kableExtra')) install.packages('kableExtra'); library('kableExtra')


url <- "https://api.census.gov/data/timeseries/intltrade/exports/hs?get=DISTRICT,DIST_NAME,E_COMMODITY,E_COMMODITY_LDESC,ALL_VAL_MO,ALL_VAL_YR,VES_VAL_MO,VES_VAL_YR,AIR_VAL_MO,AIR_VAL_YR,CC_YR,QTY_1_YR,QTY_2_YR,CTY_CODE,CTY_NAME,COMM_LVL,DF,LAST_UPDATE,YEAR,MONTH,VES_WGT_YR&YEAR=2013&MONTH=12&DISTRICT=13"

response <- GET(url, timeout(60))

Conclusion

This project aims to provide valuable insights into global trade dynamics, supporting stakeholders in making informed decisions about export strategies and market targeting.