Data Mining Consulting Services

Our data mining consulting process typically involves:

Project Assessment

Translation of business problem into solvable data mining/quantitative counterpart and evaluation of the feasibility and successfulness of a project implementation:
-    estimating objectives and goal attainability
-    determining required resources
-    checking the correspondence between objectives (business questions) and data availability
-    discussing possible findings and their usage in business decision making
-    answering the question about possible necessity of data enrichment and the usage of GIS
-    determining required data mining software.
Multidisciplinary/cross-functional project team building.

Data Preparation, Data Preprocessing and Exploratory Data Analysis
Review available data sources in light of project objectives and business questions and creating data set for the analysis. Defining / discussing targets and inputs, exploring their distributions, missing values and outliers. Visualizing the relationship between variables. Determining necessity and probing nonlinear transformations of variables.

Model Development

Formation of the adequate class of data mining models. For example, for response modeling with a binary target the following models can be considered:
- parametric logistic regression
- non-parametric models:
classification tree
neural net
stochastic gradient boosting
random forest
memory-based reasoning, etc.
Development of a predetermined subset of models. Evaluation of each model interpretability (if it is important and possible), accuracy and stability. Selection of the best model and assessment of model recommendations.     

Model Deployment

Creating software infrastructure for selected model. Monitoring model performance, and if necessary, updating/modifying the data/model.