Energy systems generate a large amount of data related to their production, sale, consumption services, etc. These numbers remain deposited into large databases with valuable hidden information about trends that would help service providers and their users make better decisions. Data science, mathematics, and statistics have been advancing to respond to these needs. This revolution has brought several techniques, algorithms, and prediction models, among which machine learning emerges. This work helps to link these tools with the decisions making of companies and users related to a Colombian natural gas producer company. The public data extracted from the company were understood, cleaned, and processed before being deployed within two predictive models: time series forecasting using artificial neural networks (TS-ANN) and k-nearest neighbors (KNN). The model targets were the fixed and variable natural gas charges regarding the demographic information of the customers. The results obtained by both models presented small deviations from the test values. Both, the KNN and TS-ANN model received the data time input for predicting natural gas charges in a selected demographic sector. As a result, we found trends to identify the consumers who receive a higher service charge and the period in which these values are higher. These results are presented as an indication of what can be done today with the public information provided by companies: allowing consumers to adjust their consumption strategies.