AI-Powered Analysis of Traditional Gold Mining in SUDAN: Unveiling Productivity, Resource Utilization, and Market Dynamics

Gold is Sudan’s most important mineral resource, and the country is one of the top producers of gold in Africa. This paper proposes Artificial Intelligence (AI) -powered analysis for building a robust Deep Learning (DL) model using an Attention-Infused Long Short-Term Memory (LSTM) network. The process begins with data preprocessing, including mining and encoding, to handle raw datasets efficiently. Following this, cleaning and preparing data ensures the removal of inconsistencies and readiness for analysis. Exploratory visualization techniques, such as heat maps and pair Principal Axis Indicator (PAI) plots, are employed to uncover underlying patterns and relationships within the data. The dataset is then partitioned into training and testing sets for model evaluation. At the core of the work is the attention-infused LSTM, which enhances the sequential learning process for improved performance. Model tuning is performed with a focus on key optimization metrics like AUC-ROC and log loss to achieve the desired accuracy and reliability. The simulation is implemented in Python software with Traditional Gold Mining in SUDAN dataset, accurately predicting an accuracy of $\mathbf{9 8. 8 7} \boldsymbol{\%}$, precision of 99%, recall of 97.9%, and AUC of 98.87% attained.