What makes iOS one of the most successful mobile operating system in the world?
An analysis to determine what Twitterati’s think about the recent iOS 14 launch.
Introduction:
An iPod, a phone, and an Internet communicator. An iPod, a phone… are you getting it? These are not three separate devices, this is one device, and we are calling it iPhone.
… said Steve Jobs when he introduced Apple’s iPhone to the world on January 9th, 2007.
The iPhone and iOS has evolved to a great extent since then.
iOS 14 - Apple’s fourteenth major release of the iOS mobile operating system was released on September 16, 2020.
Like all the excited Apple users, I was also waiting eagerly to see how Apple satisfies the feature-hungry fans’ hopes. And I must say, Apple did not let its users down.
I wanted to see what users feel after using the OS during the first few days after its release.
Twitter being the largest micro-blogging site where users post their opinions, reviews, thoughts and experiences - I could not think of a better place to start collecting data for my research.
I collected 6220 tweets within a span of three days from September 16, 2020 to September 18, 2020 related to the iOS 14 discussions and started analyzing them.
Let me take my readers on a step by step journey of the proceeding.
Prerequisite:
Twitter’s REST API allows us to read and write Twitter data(create new tweets, read user profiles and the data of the followers).
I used the open source Python package Tweepy to access the Twitter API with Python.
In order to use Tweepy, a Twitter Developer Account is required to call Twitter’s APIs.
After getting access to the developer portal, I created a developer app in order to get the API key, API secret key, Access token and the Access token secret which are required to make an API call.
Implementation:
In order to achieve the end goal, I have divided the problem into four small targets for systematic implementation. They are:
- Collecting and storing the data based on search term.
- Studying the data received for cleaning an understanding the structure.
- Predicting the data based on classifiers.
- Visualizing the prediction to on a result.
Lets go over the steps one by one separately and in succession.
a) Collecting the data:
My goal here is to collect tweets from the Twitter database using the Twitter API and identify what the majority of users think about the recent ios 14 release - Positively, Negatively, or Neutrally.
Importing Libraries:
pandas DataFrames are similar to SQL tables or the spreadsheets that we work with in Excel or Calc. In many cases, DataFrames are faster, easier to use, and more powerful than tables or spreadsheets because they’re an integral part of the Python and NumPy ecosystems.
tweepy contains the twitter API to validate the twitter account and fetch tweets from a specific hashtag.
time provides various time-related functions. I will be using it to calculate the time taken to fetch tweets and store them in csv files.
Authentication:
We are passing the twitter credentials to tweepy via its OAuthHandler.
Fetch Tweets:
Tweepy cursors fetch the requested data by performing pagination.
We need to put a search parameter such as key words and hashtags, starting date after which all tweets would be extracted, number of tweets to pull per run etc. and we will get a response back from the API.
After we have received the data, I stored them in a csv file to process them further.
b) Studying and cleaning the data:
After analyzing the data, I found non-ASCII characters present in it. Hence, I removed the non-ASCII characters from the file.
c) Predicting the Tweets:
In order to perform sentiment analysis and classify the tweets, I utilized TextBlob, which is an extremely powerful NLP library for Python. TextBlob is built upon NLTK and provides an easy to use interface to the NLTK library.
I also utilized the python CSV module to read and write the csv file containing the tweets.
For my dataset consisting of 6220 tweets, I got some fascinating results.
50% of the Tweets were positive whereas only 13% were negative Tweets.
The results clearly show the positive popularity of iOS among its users.
In order to get a preview of the predicted results I used the read_csv() function in the of the open source data analysis and manipulation tool pandas to import the predicted CSV file and preview a specified number of rows in the file. read_csv()also has an argument called skiprows that allows to specify the number of lines to skip at the start of the file.
d) Visualizing the results:
Visualization helps users to recognize relationships between the data, providing greater meaning to it.
Hence, I visualized the results using Charts so that I am able to communicate my data findings in a much easier way.
I utilized Matplotlib which is the most popular plotting library in python to create pi charts showing the percentage of Positive, Negetive and Neutral Tweets.
Next, in line was to create word clouds of the positive reviews and negative reviews separately.
By doing so, I was be able to visualize and understand the reviews from the user’s perspective about the likes and dislikes in the newly released iOS 14.
Positive prediction word cloud:
Analyzing the positive tweets, I found that words like home screen, back tap, library were some of the commonly spoken words.
This makes sense as these are words related to the new/upgraded features introduced in the iOS 14 update which are:
- Home screen widgets
- Back Tap custom control
- Updated Application Library and interface
Negative prediction word cloud:
Analyzing the negative tweets, I found that words like bug, bad, broken, were some of the commonly spoken words.
It is unfortunate that the users had faced some bugs that caused a bad experience.
Discussions:
Some of the common words that were used in both cases were widget and home screen.
Home screen and Widget being the popular additions, both have its advantages and disadvantages.
The Home Screen has been updated after many years with a smart, intuitive, modern and neat look which may be appealing to many however, it may be a bit complicated to use which makes it draw negative comments.
With the new Widget update, information can be grasped at a glance without opening the app which is a helpful feature. However handling the widgets is a tiresome task which requires a lot of effort to setup and also during using it which is not at all convenient to some of the users.
Hence, the results obtained from this analysis are justified.
Coming to our main question:
What makes iOS one of the most successful mobile operating systems in the world?
The answer is definitely its user satisfaction and belief. We were able to clearly see that the Tech Giant only got 13% negative reviews from which we can definitely say that Apple has done a great job delivering a world class software to its proud users.
Improvement Ideas:
A very interesting improvement to this project is:
I have classified a review as uniquely positive, negative or neutral. However, within one review there can be both positive and negative aspects.
I plan to take up this problem and research more about how to communicate that beyond the summary scores of sentiment prediction, if I am using a predictive model.
Conclusion:
The analysis I made is very straightforward and informative regarding the reaction of end-users on the release of the recent iOS 14 software and the dominance of Apple in the smart phone OS market. This type of analysis is very useful in case of a large datasets where it is not technically feasible to go through every tweet to detect areas of improvements for the product. However along with the advantages, a disadvantage of this analysis is that it only describes how users feel about the product and does not define a problem and how to solve it.
Twitter’s data is the most accessible and hence are largely used in academia to study topics like Sentiment Analysis, Topic Modeling, Location Analytics, Hash tag analysis etc.
Twitter made notable changes to their developer policy on Mar 10, 2020 to better support non-commercial research and further aid researchers while protecting the data from more questionable use cases.
In this project, the Twitter data privacy and other notable policies like data manipulation and handling have been followed with care and there are no ethical concerns that should be reported.
Bias in the data may arise from random selection of Hashtags and consequently missing some related hashtags. Also, the size of the data set is not sufficient to make any hard decisions for this topic. Hence, the results found are for educational purposes only.
Please reach out in case you have any suggestions or questions. I would love to connect on Facebook, Instagram and LinkedIn.
Thanks for reading!
