Since 1950, the Amazonian and Brazilian transportation complex has prioritized the model of road transport. This model does not adequately consider the singularities of each site, especially the potential for waterway transport in the Amazon. This fact probably contributed to the unsustainable local development, even though it implemented various large projects of integration, hydroelectric, mineral and agroindustry. Past studies point out that the regular roadway system that is integrated into a clandestine roadway complex is strongly related to Amazon forest deforestation. Thus, the objectives of this work are 1) investigate the influence of clandestine and regular roadway network on the deforestation of the Amazon forest and 2) to develop an approach to deforestation forecast, utilizing a time series of 30 years (1988–2018). We used machine learning in the modeling of the quantitative variables related to the transportation infrastructure, social variables and economic variables, e.g., the deforested area. The geographical study area is the state of Pará, located in the Oriental Amazon, the second largest state of Brazil in territorial extension and the most deforested. The random forest model presented the best performance with a mean absolute error (MAE) of 2534.06 and a standard deviation (STD) of 2347.67 km², pointing to a strong relationship and showing a very strong tendency. We used sensitivity analysis to evaluate the effects of regular roadway network and clandestine roadway network on deforestation. With the generated function (using the least squares method), the deforested area was estimated for the years 2020, 2030, 2040 and 2050. The results show that given the same scenario, deforestation tends to continue intensively in the next three decades. The total loss is more than 72,417.93 km² (25.77% increase compared to the current deforested area). The results show an increasing curve, although with decreasing rates of forest loss in ten years, on average from 7.80% to 0.80% per year.

自 1950 年以来，亚马逊和巴西交通综合体优先考虑公路运输模式。该模型没有充分考虑每个站点的奇异性，尤其是亚马逊地区水路运输的潜力。这一事实可能导致当地不可持续的发展，尽管它实施了各种整合、水电、矿产和农业工业的大型项目。过去的研究指出，整合到秘密道路综合体中的常规道路系统与亚马逊森林砍伐密切相关。因此，这项工作的目标是 1) 调查秘密和常规道路网络对亚马逊森林砍伐的影响，以及 2) 利用 30 年 (1988-2018) 的时间序列开发森林砍伐预测方法。我们在与交通基础设施、社会变量和经济变量（例如，森林砍伐面积）相关的定量变量的建模中使用机器学习。地理研究区是帕拉州，位于亚马逊东部，是巴西领土面积第二大、森林砍伐最严重的州。随机森林模型表现最佳，平均绝对误差 (MAE) 为 2534.06，标准差 (STD) 为 2347.67 km²、指向强关系，表现出很强的倾向性。我们使用敏感性分析来评估常规道路网络和秘密道路网络对森林砍伐的影响。使用生成的函数（使用最小二乘法），估计了 2020、2030、2040 和 2050 年的森林砍伐面积。结果表明，在同样的情景下，森林砍伐在未来 30 年中趋于密集。总损失超过 72,417.93 km ²（与目前的森林砍伐面积相比增加了 25.77%）。结果显示了一条增加的曲线，尽管森林损失率在十年内下降，平均每年从 7.80% 下降到 0.80%。