Applying Artificial Intelligence (AI) to software programming for forecasting and optimizing business management decisions
This study demonstrates that applying Artificial Intelligence, particularly Artificial Neural Networks, to software programming enables businesses to optimize management decisions and maximize ROI by accurately forecasting sales performance and identifying optimal marketing budget thresholds.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
In the modern digital economy, businesses face a constant challenge: how to spend their money wisely to grow. For decades, managers have relied on intuition and simple charts to guess how much to invest in advertising or new products. Today, however, a powerful shift is occurring. Computers are learning to find patterns in vast amounts of historical data, a field known as machine learning. This approach allows organizations to move beyond guessing and instead use simulations to predict future outcomes. The core idea is straightforward: by feeding a computer past records of spending and sales, the machine can learn the relationship between the two and suggest the most effective path forward. This is not about replacing human judgment, but about giving managers a clearer, data-driven map to navigate a volatile market where consumer habits change quickly and competition is fierce.
A recent study by Trinh Quang Minh from Can Tho University explores exactly how these computer tools can be applied to real-world business decisions. The researcher set out to test three different types of artificial intelligence models to see which one works best for predicting sales based on marketing budgets. The goal was to find the "sweet spot" for spending—the specific amount of money that generates the highest return without wasting resources. To do this, the team created a simulation using data that mimicked a mobile phone retailer in Vietnam. They fed the computer historical records showing how much money was spent on marketing and how many phones were sold in response. The data ranged from ten billion Vietnamese dong spent on advertising up to fifty billion dong.
The study compared three distinct approaches. The first was a linear regression model, a simple method that assumes sales grow at a steady, predictable rate as spending increases. The second was a decision tree, which works like a flowchart, splitting data into branches to handle sudden changes or jumps in behavior. The third was an artificial neural network, a complex system designed to mimic the human brain's ability to recognize intricate, non-linear patterns. When the researchers ran their simulations, the results revealed that the best tool depends entirely on the situation. The simple linear model worked well for steady, predictable growth, forecasting that spending fifty billion dong would lead to about 236 units sold. However, the more complex neural network predicted that the same investment would yield 240 units, suggesting it was better at capturing the subtle, messy realities of the market.
The simulation identified a specific threshold for success. It determined that spending between fifty and fifty-five billion Vietnamese dong was the optimal level to maximize efficiency. Below this level, such as spending only thirty-five billion, the models predicted significantly lower sales, indicating a missed opportunity. Above this level, the returns began to diminish, meaning the extra money spent did not generate enough extra sales to justify the cost. The decision tree model proved particularly useful for understanding sudden shifts, such as the launch of a flagship product or a major holiday sale, where consumer behavior changes abruptly rather than gradually. In these scenarios, the tree model could forecast sales of 200 units at a forty-billion-dong spend, outperforming the simpler linear model which predicted only 190.
The findings offer a clear guide for business leaders. For small companies with limited data and stable markets, the simple linear model is often sufficient and easy to understand. For businesses facing volatile markets where customer behavior jumps unexpectedly, the decision tree provides a more flexible and intuitive view. For large organizations with massive amounts of complex data, the neural network offers the highest accuracy, even though it is harder to explain how it reaches its conclusions. The study concludes that there is no single "best" model for every situation. Instead, managers must choose the tool that fits their specific data and market conditions. By making this choice, they can shift from relying on gut feelings to making highly rational, data-backed decisions that optimize their return on investment and secure their position in a rapidly changing digital economy.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.