The continuous gathering of vast amounts of environmental data from IoT sensors results in unprecedented data volumes. These data typically have high dimensionality, containing numerous attributes representing diverse sensor measurements such as temperature, humidity, and other readings depending on the specific IoT application. Transmitting these data streams to edge and cloud devices for decision support consumes significant energy and bandwidth. To mitigate this, data compression methods are employed to reduce the size of the dataset by encoding it in a more compact form while preserving its essential information. With the emergence of Deep Neural Networks (DNN), autoencoder-based data compression has gained popularity for its efficiency in compressing data. Autoencoder-based data compression, using Deep Neural Networks (DNN), efficiently reduces data size by encoding and decoding high-dimensional IoT data. We propose a method leveraging autoencoders to compress IoT data while preserving essential information. We integrate autoencoder-based data compression with a Random Forest algorithm for IoT data management. By doing so, we achieve a remarkable 68.29% more compression compared to using autoencoder alone, with minimal impact on root mean square error (RMSE). Furthermore, the integration with Random Forest enables efficient decision support while reducing energy consumption and bandwidth usage. The results demonstrate the superiority of our approach in terms of data compression, energy efficiency, and bandwidth utilization compared to utilizing autoencoder alone.