Intro

The world around us is time-dependent and seasonal, so are every aspect of our lives. Business is no exception: business analysts need predictive tools available to make sense of the time-dependent nature of the metrics they are analyzing. Forecasting is fundamental to any business strategy. A great deal of it can be done with traditional statistical methods such as times series analysis. Recently, with the advent of GPU computing and cloud technologies, machine learning has become a lot more accessible to solve such problems more accurately. Neural network-based machine learning has the advantage of being able to factor in more parameters, whether they are structured, like historical data, customer satisfaction poll numbers, macroeconomic indices, and even weather, to unstructured data such as blog posts and social media feeds. In this article, I present a few business applications in which forecasting can be carried out by machine learning or times series analysis.

1 – Forecasting the number of patients

Diseases are some of the most seasonal quantities to study. Their wild swings keep hospital staff on their toes, especially emergency staff. Being able to predict, for example, how many psychiatric patients will check in the ER on Christmas eve, allows hospitals to plan their resources accordingly and avoid emergency department overcrowding, which has been shown to affect not only patient satisfaction but the their treatment and prognosis qualities as well. Diseases, both physical and mental, are perfect examples of quantities that are affected by complex exogenous factors such as outside temperatures.

2 – Forecasting website visitors

Some of the most complex forecasting problems are website traffic. This is due to its volatility and dependency on a plethora of factors. From referral channels and their growth to social shares to the number of posts and their SEO niche, many factors have to be accounted for to produce an accurate estimate. Forecasting website traffic is not only a way to gauge future revenue, but also a way to manage resource allocation like the number of servers that are simultaneously online.

3 – Sales projections

This is probably the most common application of time-series forecasting and the most important one. Sales are the lifeblood of every business and predicting them accurately is of utmost importance. Forecasting sales allows business managers to set quotas and hence spot potential issues when they are not met. According to research from the Aberdeen Group, companies who forecast their sales accurately are 10% more likely to grow their revenue over time.

4 -Transportation forecasting

Transportation, or traffic, forecasting is the process by which we estimate the future number of vehicles on the road. City governments, navigation apps, and the Department of Transportation are just examples of entities that can benefit from forecasting traffic. Government agencies can then use the traffic projections to plan infrastructure updates and predict future issues that may arise.

5 – Forecasting stocks

This is one of the most sought after and most elusive goals in the history of business. If you have any stock investment experience at all, you certainly have heard that you cannot predict or beat the market. We yet hear of trading companies generating their weight in gold every year. Forecasting models cannot predict the market accurately, but they can in some cases predict movements (up and down) which, when exploited carefully, can bring good returns.

6 – Forecasting market share

Market share is one key strategic metric for business planning. It allows the business to evaluate its course and its performance relative to its competitors. Forecasting market share usually involves predicting the sales of the business in question, followed by the sales of its competitors, then computing the market share.

Conclusion

This is, of course, not an exhaustive list of forecasting business applications. It is rather meant as a way to demystify the topic to the newcomer or the curious mind. Other possible applications are customer satisfaction, staff turnover, demand for a product and countless others. If you have forecasting needs, get in touch! I can help you extract your data and use it to make the most accurate predictions.