The amount of total organic carbon is one of the most effective parameters in source rock evaluation. Laboratory analysis of the core and cutting samples is expensive and time consuming. To help reduce these two factors, various total organic carbon estimation models using the conventional well logs have been suggested. The suggested models are experimental and subject to local application limitations. Therefore, nowadays applying intelligent methods to estimate total organic carbon is common. In this study, 110 samples of Longmaxi formation in Sichuan sedimentary basin of china were available. After preprocessing, 106 samples have been used (85 training set and 21 test set). Three methods have been used to estimate the total organic carbon of objective erea; these methods include Multivariate Linear Regression, K Nearest Neighbor and Support Vector Regression. In all three models, the estimated results showed a good proximity to the real amounts. To reduce the costs of log running and as a result of the lack of some logs in some wells, identification of most effective logs in total organic carbon estimation was done using sensitivity analyses. These analyses include mean impact value methods, median impact value, stepwise regression, Morris method and Monte Carlo method. In the end, the results of sensitivity analyses introduced the three logs of density, neutron and gamma ray as the effective logs. A comparison was made between the results of all three models (Multivariate Linear Regression, K Nearest Neighbor and Support Vector Regression) in both cases of using all logs and using selected logs. The results showed an improved performance of models when using the selected logs; finally, the Support Vector Regression, trained using the selected logs with correlation coefficient of 0.96 on training set and 0.93 on test set, yielded the best performance.