Weploy
R&D/Machine Learning
Weploy is a leader in the recruitment technology sector, offering contingent staff via its tech platform and providing recruitment technology to other organisations. To maintain a competitive edge and leverage government R&D tax incentives, Weploy is conducting research and development projects to incorporate the latest technologies into its services.
NETWORKING WITH INDUSTRY EXPERTS
AGILE PROJECT MANAGEMENT
RECORD KEEPING
R&D REPORT WRITING
ML EXPERIMENT DESIGN
DATA COLLECTION & PREPARATION
Agile Project Management
To execute this R&D project, a dedicated subset of the product team was assembled. Each sprint, story points were allocated to the R&D tasks, with user stories and tasks assigned to team members. The R&D team maintained regular collaboration and conducted agile ceremonies to ensure the project's continued progress.
Develop a hypothesis
The formulation of a hypothesis was integral in providing a clear framework for guiding project activities. In this case, our primary aim was to construct a hypothesis focused on leveraging the data collected through our timesheeting capabilities to identify patterns and inaccuracies via a machine learning model.
Collection of records and baseline analysis
To have a dataset to work with, Weploy implemented functionality to collect data over time. This data was subsequently made available for the primary experiment and for understanding usage patterns. A substantial dataset of over 16,000 timesheet records, complemented by 570 change requests on our platform, was gathered. This dataset served as the baseline for detecting known error patterns within those records.
Prepare the data
Data labelling / preparation is an important step in developing a machine learning model. A data labelling process will be undertaken with each model attempted to determine if a trained model is able to make accurate predictions. The data was divided into 3 sets, a training set, a cross validation set and a testing set.
Run experiments testing existing ML models
In order to develop a machine learning model that can achieve the desired success criteria and comply with the governments tax incentives scheme, the team must perform experiments that test the strengths, weaknesses and performance capabilities of existing machine learning models using real-data. These insights will lead the team to developing an innovative solution model. Thus far, the team has explored classification algorithms like decision trees, logistic regression and neural networks.
As an R&D project can take several years to fully complete, the overall outcome of successfully developing an AI that can detect inaccurate timesheets is ongoing. However, as part of the governments R&D incentives and through the record keeping and reporting of all R&D activities, this project successfully enabled Weploy to take advantage of over $800,000 worth of tax incentives during this period to fund growth areas of the business.