DHAKA
DX Innovation

How TriBIG's Customized Market Recommendation AI Works: Export Market Forecasting with Cosine Similarity

The Cosine Similarity Algorithm: TriBIG's Core Engine

At the center of the TriBIG market recommendation system is a cosine-similarity-based algorithm. It converts a company's export profile and each country's market demand profile into vectors, then calculates cosine similarity between them to recommend the most suitable export markets. The model is refreshed monthly based on Korea Customs Service export data and currently provides customized market recommendations to more than 3,500 companies.

Vij
Core Formula
Market fit score
45+
Variables
Multidimensional vector
Monthly
Refresh Cycle
Customs data update
82%
Accuracy
Validated against exports
3,500+
Users
Annual corporate usage
86 countries
Coverage
Recommendation universe

Breaking Down the Vij Formula: The Market Fit Score

TriBIG's market fit score, Vij, is calculated using the following structure: Vij = (EMij × 45%) + (FMij × 45%) + 10. Here, EMij represents export market attractiveness, FMij represents company-market fit, and 10 is a base score. The final score is determined by combining multiple sub-variables under both EMij and FMij.

Components of the Vij Formula
ComponentWeightSub-VariablesData Source
EMij (Export Market Attractiveness)45%Market size, growth rate, import growth, tariff rateKorea Customs Service, UN Comtrade
FMij (Company-Market Fit)45%Product-demand matching, competitive intensity, entry barriersExport performance data, company profile
Base Score10 ptsFixed value for minimum recommendation coverage-

How the Cosine Similarity Mechanism Works

Cosine similarity measures similarity based on the angle between two vectors. In TriBIG, the system builds a company export profile vector using product mix, export scale, and lead markets, and a country market demand vector using import volumes by product, growth rates, and supplier distribution. The higher the cosine similarity between the two vectors, the more suitable that market is considered for the company. Scores are normalized on a scale from 0, meaning unrelated, to 1, meaning a near-perfect match.

Build Company Vectors
Create 20+ dimensional company profile vectors using export share by HS code, export volume, technology level, and price positioning.
↓
Build Market Vectors
Create 25+ dimensional market demand vectors by country using HS-code import volume, growth rates, tariffs, and competitive intensity.
↓
Calculate Cosine Similarity
Run batch cosine similarity calculations between each company vector and market vectors across 86 countries.
↓
Integrate EMij and FMij
Reflect cosine similarity in FMij and combine it with EMij, the market attractiveness score, to derive Vij.
↓
Rank and Recommend
Recommend the top 5 to 10 markets for each company based on Vij, along with market-entry guidance.

The Data Pipeline: Monthly Refresh

A key reason TriBIG maintains recommendation accuracy is its monthly data refresh cycle. Whenever Korea Customs trade statistics are updated, the vectors used by TriBIG are recalculated automatically. This allows market shifts, including tariff hikes, demand spikes, and the emergence of new competitors, to be reflected in recommendations without delay.

Monthly Data Inputs
Customs ExportsProduct and country performance
UN ComtradeGlobal trade statistics
FX and TariffsContinuously reflected
Company ProfilesNew and revised data
Algorithm Outputs
Market RecommendationsTop 10 by company
Fit ScoreVij 0-100
Trend ChangeUp/down vs prior month
Entry StrategyMarket-specific guide

Reading the Bangladesh Market Vector

Within TriBIG, the Bangladesh market vector reflects a high import orientation toward RMG inputs, while growth in IT and electronics is accelerating sharply. Korean exporters of garment materials, machinery and equipment, and IT hardware show relatively high cosine similarity with Bangladesh. The country is especially likely to be recommended as an alternative market for product groups exposed to elevated tariff risk in the United States and the European Union.

01
Garment Materials: Similarity 0.87
Bangladesh's import demand for inputs serving the RMG industry shows very high cosine similarity, 0.87, with Korean exporters of textiles and fabrics. Matching strength is particularly high in functional fabrics and accessories such as zippers and buttons.
02
Machinery and Equipment: Similarity 0.72
Import demand for industrial machinery and equipment is rising alongside the expansion of Bangladesh's manufacturing base. The model classifies the market as promising for Korean small and mid-sized machinery producers, with a similarity score of 0.72.
03
IT Hardware: Similarity 0.68
As Bangladesh accelerates digital transformation, imports of IT hardware and network equipment are increasing rapidly. The 0.68 similarity score indicates a meaningful opportunity set for Korean technology suppliers.
04
Alternative Market Index: Tariff Risk Included
TriBIG separately evaluates whether Bangladesh can function as an alternative production or export market for products affected by rising tariffs in the US and EU. Tariff-risk weighting is built into the Vij formula to improve the relevance of alternative-market recommendations.
TriBIG Alternative Market Discovery ManualA consolidated guide to the overall TriBIG system and how to use it
TriBIG AI Cosine Market Recommendation AlgorithmTechnical background and use cases for cosine-similarity-based market recommendation
TriBIGCosine SimilarityMarket RecommendationAlgorithmExport ForecastingKorea Customs Data
How TriBIG's Customized Market Recommendation AI Works: Export Market Forecasting with Cosine Similarity | Dhaka Trade Portal