AI - TRADITIONAL MACHINE LEARNING

What it is and the value it drives

Companies and government are trying to predict numbers all the time. Companies look to predict demand in their sales forecasts to make capacity adjustments. They try to predict prices that consumers are willing to pay, or discounts that get customers to convert. On the cost side, businesses often try to predict the price of commodities or inputs, or the right amount to spend on marketing with a return. Predicting these numbers today is often a mixture of data analysis, experience, tribal knowledge, and crowdsourcing. Some companies do better than others.


Governments also require numerical predictions. This may include everything from the number of people that may appear to 


Traditional Machine Learning is a powerful tool to predict numbers. It takes traditional transaction data, uses it to train a model, and outputs a numerical prediction. There are 2 types of numerical predictions that matter.


The first one involves predicting an actual number, for example the price of a product, or how many items of a product should be produced. This has applicability to the many financial and operational drivers of a business. Companies can use ML to predict any number of P&L drivers, or other KPIs. What drives the prediction is often a powerful set of data inputs that are beyond what a traditional team would be able to process.


The second type of prediction involves predicting Yes (1) or No (0) - done in practice by predicting the probability that something is a Yes (1) or a No (0). This has nearly boundless applicability, as businesses can ask almost limitless questions such as: will this type of customer buy? click on an ad? etc. 1 or 0 predictions are often known as classification problems and are powered by simple algorithms such as the logistic regression which outputs a number between 0 and 1. Many complex ML and DL techniques ultimately boil down to a prediction of Yes or No.


The value of this technology for companies will come from helping increase the predictive accuracy of various drivers of the business, and going beyond traditional data science to drive a quantum leap in performance.


Where it is today

Some Japanese corporations are already reaping the benefits of better numerical predictions using traditional machine learning. Japan's iconic UNIQLO worked with Google to develop Ariake - a system to predict demand. Better demand forecasts result in better production planning which reduces unwanted or unsold inventory. The system likely uses a combination of customer preferences, sales data, and other input features to predict what the actual demand may be, with some degree of confidence. [1]


Another great application in Japan has been the development of a ML model to predict energy consumption. Because energy consumption has incredible underlying costs of having running power plants, precision in forecasting kilo/ megawatts that need to be produced can make an enormous difference. The system was developed by Weather-news and uses the latest weather forecasts and energy consumption data, in order to make predictions of electrical power demand. The system was developed with the cooperation of Sumitomo Corporation, a leading Power Producer and Supplier (PPS), and Summit Energy Corporation who contributed historical power consumption data, experience in supply-and-demand planning.


A final example is SONY's work with Minnano Taxi to support taxi drivers and streamline their work. The project involved development of a demand prediction service to inform drivers of when, where, and about how many passengers are expected, within finely subdivided areas around the city. The prediction system was built by applying machine learning to the probe data of participating taxi companies, which have status data for nearly 10,000 vehicles.


Today the traditional linear, logistic, or multinomial regression algorithms to do these predictions are widely available, even "as a Service" on major cloud platforms. The way forward will depend on use case imagination and the will to build and innovate.


How it will likely evolve

Companies looking to get started in driving predictions can do so today. Those who wait, will only see this technology evolve to get easier to deploy, namely in 3 ways:


Firstly, more use cases will be shared around how to apply this technology. There is almost no limit to the types of questions that can be asked or posed, and some of the examples show that businesses need help in predicting various drivers.


Secondly, using more context data as part of the input features of these models. This can include weather data, sentiment data, traffic data or other data that can inform models of external events.


And third, leveraging AutoML models. This means ready made models as a service that only require feeding the data inputs, and leverage "pre-trained models". For example, retailers looking to do better forecasting today can get started on AWS Forecast, which has Retail forecasting, and Demand forecasting . Customers looking to answer predictive questions such as "which customers will buy" can also do this with a one line instruction on Google BigQuery. 


The key applications

As with the aforementioned examples, there are 2 broad applications. The first one is predicting numbers in the P&L, so as to get better accuracy or make better decisions. Machine Learning allows for a quantum leap beyond traditional 'data science' which relies on slicing and dicing data for insights.


The second major application is predicting operational drivers. These vary by industry sector, as all industries are looking for different outcomes. The Healthcare-Pharma industry for example may be looking to predict the probability that an R&D drug research effort works out, while an automotive manufacturer may be looking to predict the best market to launch a new vehicle. We will cover many of these examples in the industry deep dives.

Catalyzers for adoption


Online education for everyone: A unique dynamic in the world of Machine Learning is that one of its key pioneers - Andrew Ng - happens to be a gifted teacher, and run online education company Coursera. His courses have educated millions (with an M) and are an easy way to get started and build up to serious proficiency levels. The Coursera platform also has many courses from Google Cloud, AWS, and IBM. Companies at all levels need to task their digital, data science, and other specialists to pursue the various Machine Learning courses and specializations available online for very competitive costs.


Startup and Corporate collaboration: Startups doing IA-Machine Learning are getting funded in significant amounts. They build models and train them on public data. But these startups have a key challenge, which is that the data is like "oil", once it is used to train the model, its use has ended. Startups do not have constant data flowing through their models because this data comes from corporations. Startups and corporations need to collaborate quickly and openly acknowledging their advantages, one has he data, the other the expertise in improving model predictive accuracy.



Sources

[1] https://www.fastretailing.com/eng/ir/library/pdf/ar2017_en_04.pdf

[2] https://global.weathernews.com/news/13572/