This study presents the development of an electronic nose comprising eight homemade sensors with pure P3HT and doped with different materials. The objective is to electronically identify the gases exposed on these sensors and evaluate the accuracy of target-gas classification. The resistance variation for each sensor is measured over time and the collected data were processed by three different identification techniques as following; principal component analysis (PCA), linear discriminate analysis (LDA) and nearest neighbor analysis (kNN). The merit factor for the analysis is the relative modulation of the resistance is very important and computationally gives different results. In addition, the fact that we have sensors made with innovative materials where the reproducibility of the response for the same material can be a constraint in the recognition. In contrast, we have shown that despite the lack of reproducibility for the same material on two different sensors, and despite the instability during the ten last seconds, we have good recognition rates and we can even say which algorithm is better. It is noted that the LDA is the most reliable and efficient method for gas classification with a prediction accuracy equal to 100% whereas it reach 93.52% and 73.14 % for PCA and kNN, respectively for other techniques for 40% of training dataset and 60% of testing dataset.