HMICFRS EWS DEVELOPMENT COMPETENCE ASSESSMENT

Descriptions

HMICFRS independently assess the effectiveness and efficiency of police forces and fire & rescue services in England and Wales using PEEL assessments. The PEEL assessment framework outlines questions HMICFRS will actively collect, observe and assess evidence to assign a PEEL judgement HMICFRS has existing systems and functions today which use data to help support decision making and support inspections and other core HMICFRS activity like continuous monitoring assessments. A proof-of-concept Early Warning System (EWS) using Machine Learning was developed and implemented in early 2023 by HMICFRS. This predictive PEEL judgement grades for police forces on the inspection programme's 'Q5. Investigating Crime'. HMICFRS intend to develop another two questions (Q3 and Q10) with the intention of making the application of an EWS to these questions business as usual in due course. The EWS uses a set of machine learning algorithms to take published data including HMICFRS' PEEL data collection. The EWS model architecture is delivered via a Python ML environment hosted offline on HMICFRS' standalone developer laptops. HMICFRS are seeking to engage a Supplier to: - Build predictive models for Inspection questions Q3 and Q10 (in this prioritised order) using the existing architecture available to HMICFRS - If needed review best practice of ETL, modelling and output pipelines such to make it as streamlined and automated as possible. - Suggest and build a method for Q3, Q10 and the existing Q5 to be real-time i.e. not dependent on lagged data. - Produce robust documentation and quality assurance of all work completed and how it operates (for the internal data science team to operate and approve). - Work collaboratively with internal analysts, colleagues and teams to arrive at the best solution. - Explore options for using qualitative data as inputs to existing models, such as HMICFRS' EGTs and CDI assessments, including a scope and plan for implementation. - Aid and enhance the existing front-end user report built in PowerBI for new questions, utilising a user-experience centred approach.