Account Selection: Fashion Influencers on Instagram

A Streamlined Approach

To keep the scope focused, a simple yet effective method is employed for selecting influencers. It involves looking up online lists or articles featuring popular Instagram fashion influencers, like those titled “150 Top Fashion Influencers in 2023”. This strategy guarantees a wide-ranging and pertinent assortment of accounts for the dataset.

Image source: 150 TOP FASHION INFLUENCERS IN 2023 (UPDATED)

Examples of relevant web posts are listed below:

Fashion influencers’ Instagram accounts are then extracted from these articles, resulting in a preliminary list of 724 fashion influencers for dataset creation. However, the actual number of viable accounts turned out to be significantly smaller, due to some profiles being inactive or set to private at the time of data collection.

Potential Improvement

The existing methodology, while straightforward and effective, is labor-intensive and may not consistently produce a representative sample of fashion influencers.

A high-quality dataset for Instagram fashion trend analysis should ideally encompass aspects such as quantity, diversity, representativeness, balance, and timeliness.

Diversity and representativeness can be targeted through varied search queries, while quantity can be addressed through network expansion. The aspects of balance and being up-to-date will be further discussed in the Data Crawling section.

Below are several proposals for enhancing the quality of the dataset:

Diverse Search Queries

Utilizing search engines and related articles/posts is still feasible. However, the quality of search results heavily depends on the queries used. To ensure a representative set of fashion influencers, key fashion dimensions should be identified and combined in search queries.

Key dimensions might include:

Dimension	Examples
Gender	male, female, …
Age	child, teenager, 20s, 30s, …
Location	country, city, area, …
Style	vintage, street, …
Body Type	slim, plus size, …

Instead of a general search like “top fashion influencers on Instagram,” more specific searches such as “top vintage fashion influencers in New York” or “top 20s female fashion influencers” could be more effective.

Network Analysis

In addition to diverse search queries, network analysis could be employed to enlarge the pool of influencers, using the initially crawled accounts as a starting point.

Image source: DALL·E: the network of fashion influencers

The expansion process starts by crawling the followings of the seed accounts, with the rationale that fashion influencers are likely to follow other fashion influencers. After this data is gathered, a network is formed based on these following relationships. Since the seed pool consists of confirmed fashion influencers, it is reasonable to infer that the most commonly followed accounts among them are likely fashion-related.

Image source: lenamahfouf

Therefore, the top N commonly followed accounts can be selected, manually reviewed to confirm their relevance to fashion, and then added to the seed pool.

The same expansion process can be repeated with the newly augmented seed pool for several more rounds. This is done until the pool is adequately large or it is observed that the accounts added in the new round are increasingly divergent from being fashion-related.

It’s important to highlight that the manual verification process is crucial. This step ensures that only relevant and valid accounts are included in the seed pool, which is essential to maintain the quality of the expansion.

Image source: Multimodal Post Attentive Profiling for Influencer Marketing - Figure 4: Example images of Instagram posts of influencers in the eight influencer categories.

Additionally, a more contemporary approach could involve training a deep learning model to classify whether an account falls into the fashion category. An example of this method is demonstrated in the research paper Multimodal Post Attentive Profiling for Influencer Marketing. This technique could provide a more efficient and accurate means of identifying relevant fashion influencer accounts.

Implementation Details

Input

Name	Description
`influencer_patterns.json`	A JSON file containing a list of search queries for fashion influencers

Data Sample:

[
    {
        "url": "https://influencers.feedspot.com/fashion_instagram_influencers/",
        "pattern": "https://www.instagram.com/(\\w+)/"
    }
]

Process

Code	Description
`codes/crawler/crawl_influencers.py`	A Python script for crawling fashion influencers’ Instagram accounts

Output

Name	Description
`fashion_influencers.json`	A JSON file containing a list of fashion influencers’ Instagram accounts

Data Sample:

{
    "accounts": [
        "@zoesugg",
        "@princeandthebaker"
    ]
}