O1: Develop a reliable method of detecting high-confidence false-positive duplicate Place locations. A high-confidence false-positive Place location is defined as a Place with over 0.75 confidence that is more than 10m from it's "correct" location. Duplicate Place locations are defined as more than one Place locations meant for the same entity. Examples are prevalent in larger entities, such as airports.
KR1: Create an algorithm to detect false positive Place locations. Generate a dataset of 5-10k misplaced locations.
KR2: Create an alogrithm to properly distingquish most significant place in the case of duplicates.
KR3: Create a benchmark to verify accuracy of aforementioned algorithm. Confirm that >95% of dataset is accurately false-positive Places.
O2: Develop a corrective solution to accurately re-assign location coordinates of high-confidence false-positive Place locations.
KR1: Research potential approaches. Explore 3-5 different methods of correction, articulate pro's-con's of each.
KR2: Implement best corrective method found from KR1. Create a solution with scalability in mind.
KR3: Run benchmark created in O1-KR2 on corrected data. Ensure that >80% of Places in original dataset are corrected.