Model accuracy

Beyond the parameter sets, described above, a steric index, Hj, is introduced, which represents the steric hindrance of the ith atom by other atoms in the molecule. By definihon, the H value of the ith atom ranges from a minimum value of 0 to a maximum value of 1 proportional to its shielding by all other atoms of the molecule. Thus, a funchonal group next to large subshtuents will weakly contribute to the eshmated log P. Including the steric index yields a small, but significant improvement in model accuracy. 

Tables I and II present the results of the Work Group discussions for the screening and site-specific level models, respectively. The assessment in these tables is based on a ranking scale between 0 and 100 0 indicates situations where no testing has been attempted and 100 identifies areas where extensive testing has been completed with sufficient post-audits to validate the predictive capability of relevant models. The scores can also be interpreted to mean the extent to which additional field testing would improve our understanding of how well the models represent natural systems. It is important to note that the scores do not indicate model accuracy per se they show the degree to which current field testing has been able to identify or estimate model accuracy.

Cabrera et al. [50] modeled a set of 163 drugs using TOPS-MODE descriptors with a linear discriminant model to predict p-glycoprotein efflux. Model accuracy was 81% for the training set and 77.5% for a validation set of 40 molecules. A "combinatorial QSAR" approach was used by de Lima et al. [51] to test multiple model types (kNN, decision tree, binary QSAR, SVM) with multiple descriptor sets from various software packages (MolconnZ, Atom Pair, VoSurf, MOE) for the prediction of p-glycoprotein substrates for a dataset of 192 molecules. Best overall performance on a test set of 51 molecules was achieved with an SVM and AP or VolSurf descriptors (81% accuracy each). 

At the heart of the model are the heat and mass balance equations governing the chlorine gas, brine and amalgam layers within the cell as illustrated by Fig. 20.3. At a more detailed level each cell is divided into eight zones. Conditions within each zone are assumed to be constant and there is a trade-off between model accuracy and execution time associated with the number of zones. Typically eight zones have been found to be a good compromise. 

Since the uncertainty in model parameters Hmits the model accuracy, a deviation of 10% is generally considered a very good result. The measurements therefore demonstrate the validity of the presented model, in particular when considering that the temperature-dependent parameters are extracted from data that are only valid in a temperature range between 0 °C and 100 °C. 

The process of structure refinement aims to modify model parameters to obtain both the optimal agreement between calculated and observed intensities and improve model accuracy. To do so we minimise... 

These results certainly appear very promising and reassuring in terms of our computational strategies, which emphasize that rigorous validation of QSAR models as well as conservative extrapolation are responsible for a very high hit rate (Tables 16.1 and 16.2). However, additional developments of methodology are certainly required to improve model accuracy since quantitative agreement between actual and predicted anticonvulsant activity is not excellent. 

The skeletal or short mechanism is a minimum subset of the full mechanism. All species and reactions that do not contribute significantly to the modeling predictions are identified and removed from the reaction mechanism. The screening for redundant species and reactions can be done through a combination of reaction path analysis and sensitivity analysis. The reaction path analysis identifies the species and reactions that contribute significantly to the formation and consumption of reactants, intermediates, and products. The sensitivity analysis identifies the bottlenecks in the process, namely reactions that are rate limiting for the chemical conversion. The skeletal mechanism is the result of a trade-off between model complexity and model accuracy and range of applicability. 

Nonlinear GCMs For certain properties, model accuracy can be improved by including a quadratic term ... 

Model accuracy Rigorous process operations and physical properties for accuracy... 

Fuel quality, boiler loading, heat exchanger surface fouling, ambient condition, and aging of equipment will all cause the process to drift and affect the model accuracy. Adaptive tuning computations can be built in to take care of... 

The proposed model takes another approach. It was developed for multistage semibatch reactors with stationary solids and continuous co-current reactors with moving solids. It also allows for a crosscurrent stream such as gas sparged separately into any number of stages. The residence time of each stage is divided into a number of finite time intervals. Within each interval, the individual reactions are treated as successive rather than simultaneous. The model accuracy is controlled by selecting the number of intervals. 

Examination of the residual unmodelled variation in these experiments indicates that there is a nonlinearity in the relationship between the X and Y variables. This detracts from the models accuracy. To this end the inherently nonlinear capabilities of ANNs have been employed with an improved predictive capability resulting in the prediction of Fig. 11. 

These issues of model accuracy and model uncertainty were not dealt with at all in the 1960s and 1970s. They are absent from all process control textbooks, an exception being the recent textbook by Seborg et al. [13]. The ability of classical control theory to explain the demonstrated effects and difficulties is very limited. In the past 10 years, much progress has been made in the area of robust control. This progress is of obvious practical value. The main results are derived [8] and summarized elsewhere [12, 14]. 

It should also be noted that this model accuracy is distinct from the solution accuracy (or error), which is defined by... 

Thus, model accuracy has been traded in order to derive a model that is easy to interpret. 

MoQSAR represents a new way of deriving QSARs. QSAR is treated as a multiobjective optimisation problem that comprises a number of competing objectives, such as model accuracy, complexity and chemical interpretability. The result is a family of QSAR models where each model represents a different compromise in the objectives. Typically, MoQSAR is able to find models that are at least as good as those found using standard statistical methods. The method will also find models where accuracy is traded with other objectives such as chemical interpretability. When presented with the full range of models the medicinal chemist is able to select one that represents the best compromise over all objectives. 

---