A typology of 9 classes of neighborhoods was created from 18 census variables describing a neighborhood¡¯s demographic, housing, and socioeconomic characteristics. The classes were derived using a combination of a self-organizing map (SOM) and k-means clustering method. Details on the 18 comprising variables can be found by switching to ¡°Bar Charts¡± in the dropdown menu. For details on the SOM, switch to SOM on the dropdown menu and select any neighborhood on the map.
When clicked a neighborhood with the dropdown box showing ¡°Bar Charts¡±, it returns bar charts of 18 variables for this neighborhood. The values of each variable are z-scores relative to the mean and standard deviation of the metropolitan area for the selected year. Positive values are greater than the mean while negative values are below the mean. The mean z-value of each cluster is shown in table 1, and values shaded by intensity. In table 1, darker red signifies larger positive numbers, darker blue is a larger negative number. Similar color schemes are applied for bar charts. Red signifies positive values, and blue indicates negative values. Users can also hover the bar to get the interpretation of this value.
The Self-Organizing Map (SOM) is a data visualization method that arranges neighborhoods on a two-dimensional grid so that neighborhoods that are most similar to one another across the 18 input variables are located near each other on the grid. Each neighborhood is located on the grid 4 times, once for each time stamp: 1980, 1990, 2000, 2010. To visualize a neighborhood¡¯s trajectory of change, a line is drawn connecting its location. A longer trajectory generally represents a larger change. The colors on the SOM grid represent similar groupings of neighborhoods as shown in the legend of the main map. Change is also represented by the sequence the neighborhood follows through the neighborhood clusters. A national typology of neighborhood sequences was developed showing the dominant sequences of change across the United States. A neighborhood¡¯s membership to one of these sequence clusters is also shown.