Project Title: Turkish TV Series Success Prediction System

This project is expected to be implemented using the Julia or Python language.

Description

Dear Friends,

the main factors influencing the success of series are listed below:

A cast consisting of famous and well$-$known actors is crucial in obtaining the approval of TV

channels and influencing the success of the series positively, especially in reaching successful

rating thresholds by attracting the audience's attention in the early episodes.

The script is a critical success factor. Generally, the quality of the script becomes more

important than the contributions of the actors after the series starts to gain traction on the

channel and can rank highest in terms of importance.

The support provided by TV channels, such as advertising, trailers, and promotional support, is

an important factor for success. Additionally, channels can support series by choosing a day and

time with relatively lower competition $($no strong competing series$).$

The financial strength, reputation, and visibility of the production company also affect the

inclusion of strong scripts and standout actors in projects, thus positively influencing the

success of the series.

However, in recent times, series are not only produced for TV broadcast but also for various

digital platforms. Usually, series broadcast on digital platforms have fewer episodes per season

$(8-13$ episodes in some platforms, $20-40$ in others$)$ compared to those broadcast on TV $(30-40$

episodes per season$).$ This aspect needs to be considered in the model you will create.

In the TV Series Success Prediction System $($TSSP$),$ different classes of series to be identified are

given below:

A$.$ Will it be canceled in the first season? $($Unsuccessful$)|$

B$.$ Will it last only two seasons? $($Partly Successful$)$

C$.$ Will it end between three and five seasons? $($Successful$)$

D$.$ Will it end between five and ten seasons? $($Partly Mature$)$

E$.$ Will it last ten seasons or more? $($Mature$)$

For a newly started series, it is evident that attributes such as "End Year," "Number of Seasons,"

and "Total Number of Episodes" should not be known when using data mining techniques.

Therefore, it should be noted that these attributes$/$values should not be present in the test set

while evaluating the performance of the model. For each example in the test set, which is not

created by "Uniform Random Sampling," but selected with random sampling from all years, the

series that aired until the date the series started broadcasting should be used in the training set

of the model, and subsequent series should not be included. For example, if you have a series

from $2005$ in your test set, a model should be created using other series until that year, and this

test example should be tested with this model. Similarly, if there is an example from $2014$ in the

test set, a model should be created using other series until that year, and this test example

should be tested with this model. The result for each example should be recorded separately,

and the performance of the model you created should be calculated cumulatively. The

classification part of projects that only consider projects from recent years without creating a

test set selected by random sampling from all years will not be taken into account.The responsibility for creating the dataset is entirely yours. It is expected that you will perform

automatic data collection $($Web Scraping$)$ from relevant websites, weblogs, etc. to create

datasets. However, it is also expected that data from at least $25$ years, including Turkish series

that started in the $4$ th quarter of $2023,$ will be included. The relevant dataset will be delivered

with the project submission. It is recommended to have the following attributes in the collected

data. However, you can increase or decrease the number of attributes you will use. Also, it is

STRICTLY PROHIBITED to use actor rating information in both series and ultimately calculated in

the final results. The rating information can be used during the validation of the validity of the

obtained results. Additionally, sometimes the same series can be broadcast repeatedly by

different broadcasting companies. Attention should be paid to this in the model you will create.

Except for the rules specified in this document, each group is expected to determine and

implement other methods on their own.

Both recommended common attributes in the training and test sets:

Series name

Production year

Start year

Average duration

Broadcasting companies

Production company$($ies$)$

Producer

Scriptwriter$($s$)$

Music

Director

Cast

Genre$($s$)$

Original broadcasting organization$/$platform

Script

Attributes used only in the training set, unknown in the test set:

End year

Number of seasons

Total number of episodes

Was it exported or will it be exported?