International Business and Machine Learning

International Business and Machine Learning

Today traditional economic models are no more reliable predictors, any help to predict the future of trades is crucial at this time and is sought all around the world.

Businesses around the world had to update their schema due to new technology implementation, Machine Learning.

COVID has increased the attention and time of implementation of these algorithms because International trade policies have limited cross-border exchange of essential goods and that affects directly to the workforce and financial plans.

Machine Learning is a mathematical formula that tries to provide insights about the future based on experience from the past. To increase its power, the machine learning algorithm will need to use the maximum amount of data the company can provide, nevertheless, due to data protection laws and their anonymisation authorised processes this data sometimes is more limited but the access to open data has accelerated the process, for instance, Open-government data is a great fuel to power the algorithms at maximum to explain and forecast.

Different companies have different amounts of data to provide to their algorithms but there are plenty of different algorithms and combinations of them that can provide some advantages to the company that makes use of them.

There are 3 types of algorithms:

  1. Supervised. It is an algorithm that needs a training process, which means we will need to provide the data plus the answer to the question we would like the algorithm to answer in the future.
  2. Unsupervised. It is an algorithm that classifies the information and it is in the person that uses it the obligation to make the algorithm work in a useful way.
  3. Reinforcement. It is an algorithm that can repeat the experiment a lot of times and get an automatic response to the evaluation of the process done.



– Supervised algorithms can predict the price of a house with a learning process from all the sold houses in the neighborhood, or can learn how to identify an image from previous learning providing the answer to the new provided image.

– Unsupervised algorithms can classify the type of customers any shop has by the data the company has collected, for example, the amount of money spent.

– Reinforcement algorithms can learn how to win a videogame, learning from iterative games that can be played faster than humans.

As we now understand the type of algorithms through the given examples, it is likely to realise that we could mix the different algorithms in the different types of data that any company possesses or mixes them with the same data to improve the analysis, for example, to target their clients differently depending on the cluster (the group that identifies the algorithms) detected. They (unsupervised algorithm) provide a mix of our products that fit their cluster budget as other buyers from this cluster did (supervised algorithm).

Another example is the international business trade area that has evolved from short-term and subjective (straight-line) projections and medium-term (aggre-gated) projections to the use of open data from international trade transactions and commonly associated economic factors to predict future trade patterns (supervised learning) and create clustering of countries according to economic factors (unsupervised learning). Providing a range of data-driven and interpretable projections for individual commodities and trade flows that can inform policies affecting modern trade tariff reductions, in areas as diverse as services trade, competition policy, trade-related investment measures, or public procurement.



Eugenio Clavijo