Comparative Conductometric Study of Metal Halides and Nitrates in Binary Solvent Systems: A Machine Learning Approach

This study presents a comprehensive quantitative study of the conductance of metal halides and nitrates in binary solvent systems, investigating how solvent composition, electrolyte concentration, and temperature collectively affect the electrical conductivity. Metal halides, including chlorides, bromides, and iodides of alkali and alkaline earth metals, exhibit characteristic nonlinear conductivity variations arising from the differential solvation of cations and anions by binary solvent components. Metal nitrates demonstrate unique behavior due to the displaced charge distribution of the polyatomic nitrate ion and extensive hydrogen bonding capabilities in protic solvents. The experimental dataset contains forty-six observations covering metal halide concentrations from 0.1000 to 0.5900 mol/L, metal nitrate concentrations from 0.0500 to 0.3000 mol/L, solvent ratios from 18.0000 to 118.0000, temperatures from 24°C to 74°C, and resulting conductivity values from 2.7000 to 9.4000 mS/cm. Machine learning approaches, specifically slope-increasing regression and random forest regression, were used to predict solvent ratio effects on conductance measurement behavior. Both models showed exceptional predictive performance, achieving R² values of 1.0000 for the gradient boosting training and 0.9885 for the test data, while random forests achieved R² values of 0.9992 and 0.9934, respectively, confirming strong generalization capabilities and minimal overfitting.