category
Machine Learning + AI
industry
Office of Finance
technologies
Alteryx Analytic Apps
Alteryx Predictive R-based Tools
Using a low-code, no-code solution in Alteryx, we created a workflow which leveraged the optimization tool in Alteryx to perform the selection process. The workflow read in templates from an excel file to dynamically select the columns of interest, apply standard data transformations, set the appropriate constraints of values and implement a relaxed Knapsack problem to generate a set of loans which satisfies all of the business constraints
To select loans for a mortgage backed security, a Canadian financial company was looking for an automated way to select the appropriate loans. The firm needed the composition of loans to satisfy multiple business constraints such as average credit score. Currently the process was being done manually and was not a scalable process. So, the company was looking for automated ways using machine learning and analytics for a solution
The Canadian financial company was getting more concerned on the lack of scalability of their selection process for mortgage-backed securities. The manual process was causing a bottleneck in the business process and was slowing down other teams which needed the results. The company wanted to be more agile in providing the initial set of loans given new constraints for changing business needs and be more proactive in handling changes to the business.
To improve the loan selection process, the financial company understood that machine learning or analytics could help them automate the initial selection process to provide a preliminary set of loans which satisfied certain business needs. The improved speed through automation via machine learning or analytics could eliminate the bottleneck to other teams leading to improved efficiency and allow for more time to be spent on other business needs.
The team understood that they do not have all of the technical expertise to maintain a code-based solution, so they needed a simple to use tool which could implement these methods
Currently, the loan selection process for the client is done manually where an analyst selects loans and checks to see if each business constraint is satisfied. The process requires a lot of guessing and checking and is not scalable
The total size of loans and business constraints change depending on the purpose of the mortgage backed security. Due to the manual process it takes a lot of time to create each set of loans with different requirements
The manual process requires significant expertise and time to select the appropriate loans. The new process needs to be easy to use and not require much technical expertise to perform the selection process.
To help the Canadian financial company automate their loan selection process, the Compass Analytics team needed a tool that was easy to use for the end user yet had the capabilities to apply all necessary business constraints. This lead the team to develop an Analytical App in Alteryx while leveraging the Optimization tool as part of the suite of R predictive tools which Alteryx provides.
The analytical app allowed the Compass team to design a simple to use and intuitive interface for the business user to interact with. The app asks questions to the user to select template files in Excel filled with information to define the dataset to use, the business rules to apply and the target size of the loans. The template excel files follow a simple structure to fill in the appropriate information to set all necessary information to find the best set of loans.
While designing the solution, a key requirement was to make this flexible to changing requirements and could work with a different number of columns with different names and characteristics. The template excel files are used to provide the configuration information and the workflow dynamically applies the appropriate transformations to setup the data for running the optimization model.
The optimization tool in Alteryx leverages the R suite of predictive tools which allows for implementation of linear and non-linear optimization models. As part of this solution, the Compass team implemented a version of the Knapsack problem where the goal is to select a certain value of loans while following all the business constraints. The output of the optimization provides a list of loans that adhere to the business rules and add up to the specified value.
In addition to providing the selected loans given constraints, other diagnostic files are outputted to give the user context in the characteristics of the selected loans and that the composition match the expected output. An added bonus to the solution is that some characteristics of loans were desired, but the previous process only cared about satisfying the requirements. Now the solution captures all requirements and selects loans which are more desirable.
The solution developed by Compass Analytics has allowed the Canadian financial company to implement an automated loan selection process for their mortgage backed securities leading to reduced bottlenecks, an ability to quickly adapt to changing business needs and reducing the time needed to complete the process.
With the new solution, an analyst will not need to spend hours or days determining the set of selected loans by checking and guessing. The analyst will now be able to spend minutes setting up the excel template files and running the Alteryx workflow to generate the optimal set of loans. This automation significantly reduces the amount of time needed to generate the set of loans and incorporates the ability to prefer certain characteristics of loans which was missing in the manual process
A key feature of the new solution, allows for quickly adapting to changing business needs such as rerunning the selection process but having a greater total loan size, or comparing the results given slight differences in business requirements. This allows the organization to perform extensive scenario analysis on different inputs and better react to the requirements for selecting loans.
This solution allows the users to use familiar tools such as Excel to implement the template files for business constraints and scoring method and a graphical user interface to select the files to run the optimization process. The solution allows for non-technical users to use the optimization process to automatically select the loans reducing the expertise necessary to perform the selection process.