Northeastern University Using R Language with Prediction Model Questions The goal here is using R language with prediction model like linear regression to

Northeastern University Using R Language with Prediction Model Questions The goal here is using R language with prediction model like linear regression to maximize the efficiency of advertising, find the balance between cost and reaching the audiences. Attached two files are the dataset we have, and use what you need to solve the problem, and give some detailed information like why we use such and such, what is the benefit of using such techniques. Analyzing Custom Audiences on
Facebook
Project Outline
As a media and entertainment company Viacom creates content. The majority of
content gets created for, and published to, our television channels. We also utilize
our social handles on platforms such as Facebook to promote content. In order to
ensure the content reaches the right audience we can create custom audiences on
the social platforms. For example, we can create an audience composed of those
users that have engaged with a particular page. We may want to create an audience
like this so that if we have some new content that we believe users that interacted
with that page will enjoy we can target them directly. We can also create lookalike
audiences, that may have a similar social presence to a particular audience. For
example, we can create a lookalike audience for those users that have engaged with
MTV, which means the audience consists of users that look similar to those that
engage with MTV but have not engaged with MTV, this may an audience that will
enjoy content but not so far engaged with the page.
We are looking to answer a few key questions in this analysis to help determine
which and what kind of custom audiences help drive reach, and help drive
engagement. We want to determine:
• What are the attributes of custom audience that result in large reach for
minimum cost (large impressions, and low cpm)?
• What are the attributes of custom audience that result in high engagement
(large clicks and low cpc)?
• What are the attributes of custom audience that result in high efficiency for
determined by click through rate (ctr)?
• Can we build a model to predict the cpm, cpc, and ctr for a given custom
audience?
• Are lookalike audiences cheaper to target, do they have lower cpms and
cpcs? Can we summarize the difference between lookalike audiences and
other audiences?
The data
custom_audience_ads.csv
This dataset documents the custom audiences and the ads associated when a
custom audience is used for ad targeting. Every row represents an ad that uses a
custom audience, with attributes about the custom audience and the ad. If the ad_id
field is null then the custom audience was not used in ad targeting
account_id: the id of the account that owns the custom audience
approximate_count: the approximate count of the custom audience
custom_audience_id: the unique id for the custom audience
name: the name of the custom audience that provides reference on how the custom
audience was created
retention_days: Number of days to retain someone in audience
time_created: the time the custom audience was created
customer_file_source: describes how the custom audience was created
description: a description of the custom audience
is_value_based: Whether the audience is used to seed value based lookalike
data_source_type: indicate by which method the custom audience was created
data_source_sub_type: indicate by which method the custom audience was created
data_source_creation_params: indicate creation parameters by which method the
custom audience was created
ad_id: id of the ad associated with the
clicks: number of clicks on the ad
cpc: cost per click in $
impressions: Number of impressions of the ad
cpm: cost per 1000 impressions in $
ctr: click through rate, clicks per impressions in % (0-100)
lookalike_data.csv
This dataset shows attributes of custom audiences that are lookalike audiences, i.e.
they consist of users that look similar to other users.
country: the target country/countries of the audience. Countries are commaseparated.
origin_id: the id of the custom audience
origin_name: the name of the custom audience
origin_type: the type of the custom audience
ratio: the proportion to include in the custom audience. For example if the audience
name is “Everyone Who Engaged With MTV Page” and the ratio is 0.08, the
audience will consist of users who look alike to those that engage with the MTV page
with a count of 8% of the number that engage with the MTV page. So, if 1 million
people engage with the MTV page, the lookalike audience will be 80,000.
information for all other fields can be found here:
https://developers.facebook.com/docs/marketing-api/reference/customaudience/

Purchase answer to see full
attachment