How does CIC Advisor Predict Outcomes?

CIC Advisor uses several data points to create an initial model. These include, but are not limited to:

Current state of project
Historical performance data of similar projects over time
Historical performance data of resources and activities

Prediction models are built by using the following key processes:

Data collection and transformation
Data preparation
Model Selection and Training
Model Evaluation and Tuning
Prediction

Data collection is one of the most significant steps in the model building process. The quality and the quantity of the data collected in this step determines the accuracy and robustness of the prediction models.

Also, depending on the current state of the project, data distribution and data elements are factors that are considered in developing prediction models. For example, the data distribution of planned projects that are yet to commence will be different from those projects in progress. Therefore, CIC Advisor uses separate prediction models for planned projects and and in-progress projects.

For planned projects, CIC Advisor uses:

Various planned metrics such as planned duration, planned labor units, and planned cost
Schedule assessment metrics such as activities with negative lags, large durations, long lags, and negative floats

For projects in progress, CIC Advisor uses:

Various planned and actual metrics which describes how the current project has progressed so far. For example, estimated durations, cost and labor units, remaining duration, and actual duration
Schedule assessment metrics such as activities with negative lags, large durations, long lags, and negative floats
Predicted and actual delays of predecessor activities
Predicted delays of successor activities

Once the above data points are collected, they are formatted and passed through a random forest regression and classification model and an initial prediction model is developed to forecast outcomes of current projects to predict the delay likelihood. This initial model is then trained with your specific data to ensure the predictions are tailored to your organization. CIC Advisor predictions become more powerful and accurate over time through its self-learning capabilities and user feedback.

Last Published Thursday, December 7, 2023