Analyze [CITY/REGION] property tax assessment data from [YEAR] to [YEAR] to identify undervalued or overvalued properties using [MACHINE LEARNING MODEL TYPE]. Incorporate features like [PROPERTY CHARACTERISTICS], [LOCATION FACTORS], and [MARKET TRENDS]. Provide a ranked list of top [NUMBER] properties with the highest potential tax savings, along with confidence scores and actionable recommendations for appeals or reassessments. Highlight any systemic biases in the assessment methodology.