In oil and gas field, for emission inventory estimation purpose, we had predicted gas composition and also measured natural gas for about 10 components (30 times). For validation of estimated data, there is a dataset with 30 replicate for two columns (10 components measures for 30 times + 10 components predicts for 30 times). Now, What is the best statistical method for comparing between predicted results and actual measured?
Dear Seyed,
In statistical method, if your target equation can be shown by general equation Y=f(Xi), (where Xi - experimental factors/terms and Y - experimental results) then you can apply RSM (Response Surface Methodology), ANN (Artificial Neuron Network), ANN combined GA to optimize experimental conditions and determine the difference between predicted results and actual measured. There are many references about these methods on RG website as well as google.
Good luck
A confusion matrix would also work to assess the predicted vs actual results without relying on a simple accuracy assessment.
Dear Dr. Ghasempouri, Salam, As you know better than me, several statistical methods can be used, from the simplest ones like regression to more complicated methods such as ANN or ANFIS. I've used ANN/ANFIS for comparing predicting pathogens in food (It has been published recently in Food Control); however ANN has several disadvantage, it needs a huge dataset and has a black-box limitations. If you want to check the accuracy of prediction and validate the model based on observed data, I suggest the Mean Sum of Squared Error, Root Mean Sum of Squared Error and R2 which widely employed for the validation and fitness of models. These 3 statistical methods can be calculated easily with the help of excel. (I can provide you a template in excel ). Best regards, Esmail
Dear Prof. Ghasempouri,
If you want a systems integrated model of mining with a "Nobel" Blasting potential) so that it applies to coal mining as well as oil and gas mining, a simple model taking care of complexity and information and energy conservation simultaneously is developed as a simple Dbranes String Theory application in which we derive the Econophysically efficient "Beehive Structure" as opposed to "Tunnel Community structure" and also the Genetic Algorithm of mining which bears a certain group homology with the environment of the problem spacextime thereby resultant of the String Theory forces conformal flows result in spacextime integration prserving genetic quality and the Arrow of Time and the Hamiltonian operates on a mreanfield, thereby canonically restricting backward flow and forward dump accumulation and social welfare of the man-machine potentially Quantum System. It is therefore by Fayes logic potentially embedded in Higgs-Englert-Boson Fields of electron-proton-graviton flows particle systems and have mechanical levers of control like the diameter-depth ratio. (Mallick (2014,2015), Mallick, Hamburger, Mallick (2016)).. Which supports the conformal flows by experiments. I hope it answers your question partially.
Soumitra K. Mallick
for Soumitra K. Mallick, Nick Hamburger, Sandipan Mallick
- Badges
- Science topic
- Similar topics
- Engineering
- Petroleum Engineering
- Gas