“Hot street” of crime detection in London borough and lockdown impacts

Figures & data

Figure 1. Crime rate in Haringey compared to London in 2020.

Figure 2. Haringey’s crime distribution and land use distribution.

Table 1. Information of features.

Feature	Data Source
Street length, Carriageway width	Haringey street map
Park, Pub, Bus stop, CCTV cameras, Street lighting, Traffic signals, Zebra and pelican crossings [Note: These features refer to their counts.]	OpenStreetMap, the Haringey government
the Index of Multiple Deprivation (IMD), Average house price, Average household income, Mobility change	Ministry of Housing, London Datastore, GoogleCommunityReport,Open Street Map

Figure 3. Flowchart of the crime hot street map.

Table 2. Comparison of spatial regression models.

Models	Formula
Spatial Lag	${\gamma }_i=\mathit{\rho }\sum _{\mathit{j}=1}^n{W}_{\mathit{ij}}{\gamma }_j+\sum _{\mathit{q}=1}^Q{X}_{\mathit{iq}}{\beta }_q+{\epsilon }_i,\mathit{i}\ne \mathit{j}$
Spatial Error	${\gamma }_i=\mathit{\lambda }\sum _{\mathit{j}=1}^n{W}_{\mathit{ij}}{\epsilon }_j+\sum _{\mathit{q}=1}^Q{X}_{\mathit{iq}}{\beta }_q+{\mu }_i,\mathit{i}\ne \mathit{j}$
Geographically Weighted Regression	$Y_i\left(\mathit{u},\mathit{v}\right)={\beta }_0\left(\mathit{u},\mathit{v}\right)+\sum _{\mathit{k}=1}^p{\beta }_k\left(\mathit{u},\mathit{v}\right)X_{\mathit{ik}}+{\epsilon }_i\left(\mathit{u},\mathit{v}\right)$

${\gamma }_i$, $Y_i$ are the dependent variables in area $\mathit{i},$ $W_{\mathit{ij}}$ is the n × n spatial weights matrix of the features in area $\mathit{i}$ and its neighborhood $\mathit{j}$ , $\mathit{\rho }$ and λ are the spatial autoregressive parameters. W_ijy_j is the “spatial lag” of ${\gamma }_i$, $W_{\mathit{ij}}{\epsilon }_j$ is the values of the residuals in neighboring locations, also namely “spatial error”. $X_{\mathit{iq}}$ is an observation of the explanatory variable, Q is the number of explanatory variables, ${\beta }_q$ and ${\beta }_k$ are the corresponding regression coefficients, ${\epsilon }_i$ and ${\mu }_i$ are the error terms. (u, v) is the coordinates, $X_{\mathit{ik}}$ is the observed value of the explanatory variable X_k at the coordinate (u, v), ${\beta }_0$ is the geographically weighted expression.

Figure 4. Number of crimes in Haringey from 2017.01 to 2020.12.

Figure 5. The results of global Moran’s I.

Figure 6. The results of local spatial autocorrelation.

Figure 7. Three-Dimensional visualisation of crime density in Haringey (2020 compared to 2019, zoom in on the value at a scale of 1 mile).

Figure 8. “Hot street” maps of different crimes in Haringey in 2020.

Figure 9. Crime change of different crimes in Haringey (2020 VS 2019).

Table 3. Haringey’s number of street segments by different crime types.

Crime Type	Crime-free	“Hottest Street” of Crime
Property crimes	461	1116
Anti-social behaviour	680	1061
Violence and sexual offences	619	1076
Drug crimes	1845	770
Public order offences	1697	807
Possession of weapons	3767	289

Table 4. Comparison of the results of four regression models.

Ordinary Least Squares Regression
Sum squared residual:		8852.566
Mean dependent var:	0.8495	R-squared:		0.1689
S.D. dependent var:	1.3042	Adjusted R-squared:		0.1679
Akaike info criterion:	19958.923	Schwarz criterion:		20,019.605
Maximum Likelihood Spatial Lag
Mean dependent var:	0.8495	Pseudo R-squared:		0.6184
S.D. dependent var:	1.3042	Spatial Pseudo R-squared:		0.0023
Akaike info criterion:	16,144.448	Schwarz criterion:		16,211.872
Maximum Likelihood Spatial Error
Mean dependent var:	0.8495	Pseudo R-squared:		0.1653
S.D. dependent var:	1.3042
Akaike info criterion:	16235.083	Schwarz criterion:		16,295.765
Geographically Weighted Regression
Spatial kernel:	Adaptive bisquare Bandwidth used:			59.000
Residual sum of squares:		1 602.686
R-squared:	0.744	AIC:		11,483.712
Adjusted R-squared:	0.688	AIC c:		11,975.162

Table 5. Top 10 “Hot street” of crime in 2020.

Property Crime	Anti-social Behaviour	Violence and Sexual Offence	Drugs	Public Order	Possession of Weapons
High Road (Wood Green)	Factory Lane	Reform Row	Neville Place	Reform Row	Fitzwarren Gardens
Cannon Road	Stoneleigh Road	Factory Lane	Bennetts Close	Dowsett Road	Grovelands Road
Factory Lane	Wilson’s Avenue	Dowsett Road	High Road (Wood Green)	Factory Lane	Bennetts Close
Reform Row	Dowsett Road	Hale Gardens	Cooperage Close	High Road (Tottenham)	The Hale
Dowsett Road	Honeysett Road High Road	The Hale	Roebuck Close	Stamford Hill	Hale Gardens
Gladstone Mews	(Bruce Grove/Tottenham)	Nevile Place	Northumbert and Park	Bennetts Close	Hale Road
Pelham Road	Bramble Close	High Cross Road	Portree Close	Portree Close	High Cross Road
Bennetts Close	Reform Row	Cannon Road	The Hale	Roebuck Close	Monument Way
Stoneleigh Road	White Hart Lane	Bennetts Close	Roebuck Close	Cooperage Close	High Road (Tottenham)
Ferry Lane	Bruce Grove	Monument Way	The Avenue	Neville Place	West Green Road

Figure 10. Estimates of parameters and R-square of streets in Haringey estimates of parameters and R-square of streets in Haringey.

Figure 11. Haringey’s mobility change in April, August, and November 2020.

Figure 12. Crime changes in Haringey (2020 VS 2019).

Data Availability Statement

Data available on request from the authors. The data that support the findings of this study are available from the corresponding author, upon reasonable request.