Adaptive learning rate schedules are the most crucial hyper-parameters, which aim to automate the tedious and exhausting process of manually determining an appropriate learning rate. This paper combines a hybrid learn rate schedule method with one or more learn rate schedules with hybrid learn rate scheduling techniques to train Deep Learning (DL) models. To solve problems using one or more learn rate schedules, the hybrid learning rate schedule mechanism modifies the learning rate for every step or iteration in DL models and optimizers. To develop an efficient hybrid learn rate schedule for deep learning models, this study combines adaptive learning rate mechanisms such as piecewise, exponential decay, polynomial time, reciprocal time, and cosine annealing decay with DL models and optimizers like Adadelta, Adam, RMSprop and Stochastic Gradient Descent with Momentum (SGDM) to increase the precision of localization using the Received Signal Strength Indicator (RSSI) in Internet of Things (IoT) networks based on Long Range Wide Area Networks (LoRaWAN). The hybrid learn rate schedules with transitions, weighted averaging, adaptive blending, harmonic mean, cyclic blending, and geometric mean enhance the model’s performance by adding flexibility to the learning process and deviating from traditional predefined learning rates. To map with the LoRaWAN RSSI-based localization datasets and retrieve the performance parameters, the mathematical model of the hybrid learning rate schedule is constructed and formulated using adaptive deep learning rate models. The DL models’ default parameter values results are compared for each possible hybrid learning rate schedule technique, and the results are higher and more accurate than those of the current models.