Applying Artificial intelligence and Big data analysis in predicting the effectiveness of Strategic decision-making in the Hotel Industry - Decision Support System

In the world of high-stakes business, the difference between a thriving enterprise and a failing one often comes down to a single, critical choice: where to put the money. For decades, leaders in the hotel industry have made these strategic decisions based on experience, gut feeling, and reports that look backward at what already happened. They would decide whether to build a new wing, restructure a debt, or change their pricing strategy by reviewing past financial statements and hoping the future would look similar to the past. However, the modern market is too fast and too complex for this old approach. Consumer tastes shift overnight, economic conditions fluctuate wildly, and the sheer volume of information available is overwhelming. To navigate this, a new approach has emerged that treats business strategy less like a gamble and more like a science, using vast amounts of data to predict what will happen before a single dollar is spent.

This shift relies on two powerful tools working together. The first is big data, which is simply the collection of massive amounts of information from many different places, such as official financial records, industry reports, and even the unstructured text of online customer reviews. The second is artificial intelligence, specifically a type of computer program called machine learning. Unlike traditional software that follows rigid, pre-written rules, machine learning allows a computer to study patterns in data on its own. It learns to recognize which combinations of factors—like how much debt a company has, how much profit it made last year, and how happy its customers are—tend to lead to success or failure in the future. By feeding these systems historical data, researchers can train them to forecast the outcomes of future decisions with a high degree of accuracy.

Trinh Quang Minh, a researcher at Tay Do University, set out to test whether this combination of big data and artificial intelligence could actually help hotel executives make better strategic choices. The study focused on the hotel industry in Vietnam, a market known for its rapid growth and significant volatility. The researcher gathered a decade's worth of information, spanning from 2010 to 2025, from a wide variety of hotel companies. This included detailed, audited financial reports from major groups like Saigontourist and Vinpearl, as well as data from smaller, listed companies and boutique hotels. The data was not limited to numbers; it also included industry benchmarks like average room rates and occupancy numbers, as well as the raw text of online customer reviews to gauge public sentiment.

The core of the research involved building a digital engine capable of simulating different business scenarios. The researcher took the collected data and cleaned it, filling in missing gaps and organizing the numbers so they could be compared fairly across companies of different sizes. Then, they trained several different types of machine learning models to act as predictors. Some of these models were traditional statistical methods, while others were more advanced, tree-based systems that could handle complex, non-linear relationships between variables. The goal was to see which model could best predict whether a specific strategic decision, such as a capital investment or a restructuring plan, would lead to a financial recovery and long-term success.

The results indicated that advanced machine learning models, particularly those based on tree-structured ensembles like XGBoost and Random Forest, significantly outperformed traditional statistical methods in broader benchmarks. In tests designed to forecast multi-period financial recovery, the most sophisticated model achieved an accuracy of over 91 percent. However, the study explicitly warns that when the model was tested on a very small, specific sample set (only six samples), it achieved 100% accuracy, a result the author attributes to "overfitting" and high data separability rather than true generalization. The paper notes that while the models were better at spotting non-linear connections between business factors than older linear methods, the perfect scores on the small test set were likely due to the limited sample size, and large-scale implementation requires further data to optimize generalization capabilities.

Beyond just predicting success or failure, the research also uncovered exactly what drives those outcomes. By analyzing which pieces of information the computer relied on most heavily, the study identified the dominant factors behind strategic success. The most critical drivers were pre-decision financial health, specifically how much debt a company carried relative to its assets and how much it was spending on capital projects. These financial metrics accounted for more than a third of the predictive power. The second most important factor was the company's historical operational performance, such as its gross operating profit and revenue per available room. Interestingly, the study also found that unstructured customer sentiment—the feelings expressed in online reviews—played a vital role, acting as a market moderator that could influence how quickly a hotel recovered after a major decision.

The ultimate goal of this work was not just to build a better prediction tool, but to create a practical decision-support system for hotel boards and executives. This system allows leaders to run "what-if" simulations before they commit to expensive strategies. They can ask the system questions like, "If we take on this much debt to build a new wing, how will it affect our profits in three years?" or "If we change our pricing strategy now, how will customer sentiment impact our occupancy?" The study suggests that by using this data-driven approach, hotel leaders can move away from relying on intuition and instead make high-stakes decisions with a clear, evidence-based understanding of the risks and potential rewards.

While the findings are promising, the researcher acknowledges that the study has limitations, primarily due to the size of the dataset used for the final testing. The models were trained on a specific set of Vietnamese hotel data, and while the results were highly accurate within that context, the researcher notes that future work should expand to include a broader range of hotels across different regions to ensure the predictions hold true everywhere. Despite this, the study provides a strong foundation for a new way of managing hospitality businesses. It demonstrates that by combining historical financial data with modern artificial intelligence, the hotel industry can gain a powerful new lens through which to view its future, turning uncertainty into a manageable variable and strategic planning into a precise science.