Food Safety in the UK: Analyzing Trends with Google Data and FSA Data

This report investigates food safety trends in the UK using Google Trends data from January 2016 to December 2018, alongside data from the Food Safety Authority (FSA). The study employs time series analysis to examine search patterns related to food quality, assured food standards, food safety, and food processing. The results reveal a general increase in Google searches related to food safety, suggesting growing public awareness. Furthermore, the analysis of FSA data indicates the public's increasing reliance on online sources for information. The report includes visualisations of the trends, decomposed time series data, and an analysis of information requests made to the FSA. The findings highlight evolving consumer behaviours and the potential of online data to inform policy and research in the realm of food safety.

Insights of Food Safety using Google Trends 1
Insights of Food Safety using Google Trends
Name
Course Number
Date
Faculty Name

Insights of Food Safety using Google Trends 2
Abstract
There are lots of data that are recorded because of human-technology interactions and they can
be used to predict patterns. This paper focuses on acquiring insights on food safety by using
google trends data and linking the findings with information sharing trends from the Food Safety
Authority of the UK. Time series analysis strategies are adopted in the google trends data. The
study found that there was a general increase in the number of google searches on food safety
after analysing data from 2016 to 2018. The study results mean that people are becoming more
sensitive about food safety. In addition, there is an increasing number of people opting to seek
food safety information from online sources. Therefore, online data can be used to extract human
behaviours and hence inform better policies and research targets.
Insights of Food Safety using Google Trends
Introduction
Food safety is crucial for the health of residents and the mandate of controlling foods entering
the market of a county is with the state department for agriculture. In the UK, the Food Safety
Authority (FSA) is in charge of all the food quality and safety concerns. The FSA develops
policies which ensure that the citizens and residents within the country get access to safe and
quality foodstuffs regardless of whether is fresh produce, cereals or servings in a restaurant. In
collaboration with the bureau of standards, they make sure that any imports concerning foods
meet the policy criteria. The FSA has created the freedom to information policy and the general
public and institutions can enquire information concerning food safety. The behaviour of seeking
information would not possibly be constant all year. Issues relating to food safety would most
likely be controlled by the media or events concerning food poisoning or the possibility of
having low-quality foodstuffs in the market.
In the contemporary world, people are keen on the kinds of food they consume due to emerging
linkages between non-communicable diseases and lifestyle behaviours. The trends of seeking
such information by the general population would also be influenced by the recent discussing the
public domain or case scenarios such as published articles in the discussion. Currently, human
and technology interaction leads to the accumulation of a lot of data that can be used to draw
meaningful insights about the prevailing behaviours (Sivarajah et al., 2017). Most of the

Insights of Food Safety using Google Trends 3
withstanding world problems could probably be solved by the big data such as Facebook, Google
and Twitter among other technology interactions (Hoffmann, 2013; Merino et al., 2016). Online
data can also be linked with offline data to detect patterns to build stronger evidence (Kosinski et
al., 2016).
Google trends data on searches regarding food safety, quality, assured food standards, and food
processing will be used to set out meaningful patterns. The Google trends data will also be
analysed in line with data in freedom of information logs from the FSA. The data contains
information on how people sought information from the FSA. Since it is an association in charge
of food safety, the study assumes that all the information sought was directly related to the
quality and safety of foodstuff within the UK. From the FSA freedom of information dataset,
information on how long it took to avail the information will be used as an indicator of the
willingness of the association to share information on food safety. It is expected that the
association takes less time to avail information to the general public as it does to agencies and
institutions. The delay for availing information to the general population might be due to
information curation and control of possible rumour to the general public about food safety.
Methods and Results
Line charts are used to visualize the Google trends data for the 4 indicators of food safety
obtained for the period from January 2016 to December 2018. The data has 5 variables, week,
food quality, assured food standards, food safety and food processing. The week variable is a
date variable aggregated as a week counts. The Google trends were extracted as google searches
but not as topics. Some observations had zero counts, meaning that in a particular week had no
searches among the four was done via Google. These kinds of observation were filtered from the
dataset using the filter_at function in the data management dplyr package. The function was
implemented by specifying the four variables counting counts and applied criteria of removing
observation were all observations equal to zero. The variables where the filter criteria are applied
are selected using the .vars argument within the filter_at function and specifying their column
indexes within the column. After filtering the observation with zero counts across the variables,
they reduced from 157 to 130.
Transforming data from wide to long format is important to make good visualisations in R. The
data was transformed using gather function in the tidyr package (Kabacoff, 2011; Chang, 2013;

Insights of Food Safety using Google Trends 4
R_Studio, 2015). The new dataset contains three variables; week, a date variable, Google search
a categorical variable with the searches included in the Google trend data extraction and finally
the count which is aggregated per week. Determining patterns in a dataset would require
choosing an appropriate data visualisation tool and manipulation the data presentation to
maximize the observed patterns. In this analysis, geom_smooth function implemented in the
ggplot package was used. The functions implement local regression (loess), a form of non-linear
regression. Span argument within the package, which takes a value between 0 and 1, can be used
to determine the length of the independent variable measure to adopt in applying the loess
regression. For the case of this dataset, a span of 0.1, meaning short distance is adopted, and the
time series visualisation depicts pattern changes.
0
10
20
30
40
Time in Dates (by 2 months)
Number of google searches in a week
assured_food_standards food_processing food_quality food_safety
Figure 1: Time series plot for the number of google searches related to food safety in the UK
between January 2016 and December 2018. The x-axis shows the time variable with date value
2-month breaks. The continuous show the time series data smoothed using a span of 0.1 and
dashed line using the stat_smooth default value of 0.75.
Figure 1 shows that there was variation in the number of google searches related to food safety
between 2016 and 2018. A low number of googles searches on assured food standards are
observed with some spikes being observed one in years. In 2016, a significantly higher number
of searches on assured food standards were recorded around April and this was replicated in
2017 but a similar pattern is not observed in 2018. A similar pattern was observed in a time

Insights of Food Safety using Google Trends 5
different period in 2018 which was around October. In all the other months, there are
approximately zero searches on assured food standards.
A lot of variation is observed of food processing, quality and safety searches. There is a general
increasing trend in food processing searches in 2016, which decreases in 2018 and stabilizes in
2019. Across 2016, there is an averagely stabilized number of searches on food quality, which is
around 25 searches a week, while a decreasing trend is observed in a greater part of 2017 and an
upward trend observed in 2018. On average, there is a stabilized number of searches per week on
food safety between 2016 and 2018. However, a slight increase in the number of food safety
searches is observed from around November 2017.
dataseasonaltrendremainder
2016 2017 2018 2019
0
25
50
75
100
-20
0
20
17.5
20.0
22.5
25.0
27.5
30.0
-50
-25
0
25
50
Time
Food safety decomposed time series
Figure 2: Line plots for decomposed time series of food safety searches. The x-axis is the time
variable cutting from January 2016 to December 2018. The data plot is a time series plot for the
weekly aggregated number of Google searches. The seasonal plot shows the season variation
observed in the data. Trends is the decomposed trend from the data. The remainder is the plot of
the unaccounted variation in the data.

Insights of Food Safety using Google Trends 6
There is no much variation observed on the week aggregated number of searches on the assured
food standards. Therefore, its time series was not decomposed. The decompose function in stats
package was used to extract the seasonal, trend and random components with the time series
data. Further, the resulting time series objects were visualized using autoplot function in ggplot2
package. The seasonal plot in figure 2 shows that there is seasonal variation where higher google
search counts on food safety are observed in the first half of the year and lower counts in the
later second half. The decomposed trend shows that there is a general increasing trend in the
aggregated weekly number of counts. The remainder plot shows a lot of unexplained variation
(Siegel, 2012).
dataseasonaltrendremainder
2016 2017 2018 2019
0
20
40
60
80
-20
0
20
40
12.5
15.0
17.5
20.0
-40
-20
0
20
Time
Food quality decomposed time series
Figure 3: Line plots for decomposed time series of food quality searches. The x-axis is the time
variable cutting from January 2016 to December 2018. The data plot is a time series plot for the
weekly aggregated number of Google searches. The seasonal plot shows the season variation
observed in the data. Trends is the decomposed trend from the data. The remainder is the plot of
the unaccounted variation in the data.
The time series on weekly aggregated google searches on food quality was decomposed and
visualized as shown in figure 3. The data plot shows the time series plot on the number of

Insights of Food Safety using Google Trends 7
searches. There is seasonal variation showing that slightly higher counts are observed in the first
quarter of the year. An upward trend is observed in the second half of 2016 while a decreasing
trend is observed for the entire 2017. The first half of 2017 show a slow decrease in trend while a
sharper decrease in a number of food quality searches is observed in the second half. In 2018, the
number of searches increases significantly depicting an upward trend. There is a lot of
unexplained variation which is shown in the ‘remainder’ plot in figure 2. In a similar analysis, an
upward trend for the food processing searches are observed in the second half of 2016, an overall
decreasing trend in 2017 and a stabilized trend after around December 2018. There is also a
significant amount of unexplained variation of the number of food processing google searches as
shown in figure 3, remainder plot.
10
15
20
Month
N u m b e r o f in fo rm a tio n re q u e s ts
2016 2017 2018
Figure 4: Time series plots for the number of information requests at Food Safety Association
for the years 2016, 2017 and 2018 coloured red, blue and green respectively. The dashed lines
show smoothed loess regression with a span value of 0.7 and 0.2 for the continuous lines. The x-
axis is a time variable showing the months and y variable is the number of information requests
aggregated per month.
In figure4, the number of information request aggregated per month are depicted using line plots.
On average, the number of information requests is lower in the first half across the three years
while more request is observed for the second half. Stabilized trends are observed in the first half
of the years while a steep increase in the number of requests is observed between June and

Insights of Food Safety using Google Trends 8
September for 2017 and 2016. An increasing trend is observed later in the year between August
and December.
Further did show that most of the information request was from the members of the public. This
result shows that the members of the public are highly interested in information regarding food
safety and quality. Media follow as number 2 of the most source of information request at the
FSA. Others include individuals from legal, business, academia, and pressure group among
others. There are no much differences in the number of days taken before responding to the
requests. Most of the responses were provided within 20 days after the request. Some few cases
for requests emerging from industry, legal media, pressure groups and the general public took
more than 40 days before the response. For the information requested by either of the categories,
not all the information was provided. Information requests in the business category show the
highest proportion of all information provided. The fewer proportion of all information provided
is observed for requests from other public authority and those from legal.
Discussion
The number of google searches on assured food standard was found to significantly low across
the three years. However, some spikes were observed at a similar time on 2016 ad 2017 while a
different time point was observed for 2018. This result shows that fewer people are interested in
understanding whether standards of food are assured. The different spikes in the number of
individuals searching assured food standards would be influenced by localized phenomena
touching on food quality. Therefore, seasonal variation is observed for the 2016 and 2017 but
including 2018 data to the analysis, seasonal is masked by random variation.
Food processing is another topic people might be triggered to search in cases of reported food
poisoning or low-quality distribution or connection between processed food and non-
communicable diseases. Looking into the google trends for this google search, ‘food processing’,
it was observed that there was a decreasing trend for the year 2017. In 2016, an increasing trend
was observed indicating that there were several discussions going on concerning food
processing. By the end of 2018, the number of google searches had stabilized to around 10
google searches in a week. After time series analysis for the food processing data, it was
observed that the trend was not consistent across the 3 years. The patterns show that most of the

Insights of Food Safety using Google Trends 9
discussions about food processing were trending later in 2016 and early 2017. After that, a
decline in the number of google searches recorded was observed at stabilized henceforth.
Food quality was another marker to the engagement on the general public on matters concerning
food safety. The trends of google searches on the term ‘food quality’ did show a lot of variation.
After decomposing the time series components, it was observed that a stabilized trend was
observed between January 2016 and March 2017. A decreasing trend was observed after June
2017 up to the end of the year and thereafter an increasing trend. This result shows a varying
trend for google searches on food quality. A lot of variation on the food quality was unexplained
and this could be due to other possible sources of variation such as trending topics or local
occurrences which might trigger the decision to look for more information.
Finally, food safety is another topic of interest which most of the people would think of search
when an issue of food poisoning emerges. Also, it is a more direct and easy to understand clause.
A time series plot did show that google searches on food safety remained significantly high
across all the three years, from 2016 to 2018. Also, compared to the other searches, food quality,
food processing and assured food standard, food safety searches were significantly high. A
slightly stabilized trend was observed from 2016 to around January 2017, after which the number
of searches started to increase. Therefore, an overall increasing trend of the number of google
searches on food safety is observed. This increase shows that people are becoming more aware
of foodstuff and are increasingly seeking information about food safety.
The google trends data was linked to freedom of information data log from the FSA. We
analysed the data and observed that most of the information request made were from the general
public. There was no significant difference in time take to respond to the requests. However,
significant differences in the proportion of providing all the information requested. Similar
trends in the number of information requests were observed. The analysis showed an increasing
trend on the number of information requests although a decline was observed between October
and December in 2018.
In conclusion, the kind of information produced by the online was also observed on offline data.
These results indicate that online data can be used to predict and detect human patterns and
behaviours. Therefore, online data is useful in providing insights for food safety based on the
analysed performed on google trends data.

Insights of Food Safety using Google Trends 10
References
Chang, W. (2013) R Graphics Cookbook Practical Recipes for Visualizing Data, World.
Hoffmann, L. (2013) ‘Looking back at big data’, Communications of the ACM. doi:
10.1145/2436256.2436263.
Kabacoff, R. I. (2011) R IN ACTION: Data analysis and graphics with R, Online. doi: citeulike-
article-id:10054678.
Kosinski, M. et al. (2016) ‘Mining big data to extract patterns and predict real-life outcomes’,
Psychological Methods. doi: 10.1037/met0000105.
Merino, J. et al. (2016) ‘A Data Quality in Use model for Big Data’, Future Generation
Computer Systems. doi: 10.1016/j.future.2015.11.024.
R_Studio (2015) ‘Data Visualization with ggplot2: Cheat Sheet’, Cheat Sheet. doi:
10.1080/15228959.2015.1060147.
Siegel, A. F. (2012) Practical Business Statistics, Practical Business Statistics. doi:
10.1016/B978-0-12-385208-3.00014-6.
Sivarajah, U. et al. (2017) ‘Critical analysis of Big Data challenges and analytical methods’,
Journal of Business Research. doi: 10.1016/j.jbusres.2016.08.001.