هوم ورک دانشجویی ماشین لرنینگ با استفاده از clustering

---

شرح پروژه

Task:

In this folder you will find a feather file nuts.feather (you may need to install a missing library with pip install pyarrow to load it) with almost 2200 data sets containing geometry data from nuts. The nuts come from several types of nuts, which and how many is unknown.

Please implement a mini machine learning project that uses clustering to try to clarify the number and affiliation of/to types of nuts.

Please note the following notes:

Project phases:

-> Data understanding (inspect data, load) -> Hint: Check in particular whether you need all columns or whether columns "depend" on each other

-> Visualization: Can I see connections from the diagrams?

-> Data cleansing (clean data, deal with quality problems) -> Hint: there will be very little to do here

-> Dimension reduction, clustering, validating results: Attention, it might make sense to use 3 principal components instead of 2 (as is often the case). Argue why, or why not. Please also pay attention to scaling!

Code snippet for 3D plot (based on our US crime data):

kmeans = KMeans(n_clusters=3, random_state=42).fit(dimred_USA)

fig = plt.figure(figsize=(10, 6))

ax = fig.add_subplot(111, projection="3d")

fig.add_axes(ax)

ax.scatter(dimred_USA['Z_1'], dimred_USA['Z_2'], dimred_USA['Z_3'], c=kmeans.labels_, cmap="viridis", s=5, alpha=0.8)

ax.view_init(azim=-66, elev=12)

plt.show()

plt.close()

Please deal with all project phases in detail and leave comments on them in english.

Determine accuracy based on the solution: The visualization of the clusters will probably not show whether the recognized nut types (or their number) are plausible. There is a second file for this purpose: nuts_solution.feather. This contains the initial data plus an additional field "Class". This contains the nut type (= label). Please calculate the accuracy with which you correctly recognized the nut types. Tune your clustering for optimal accuracy (at least 85% accuracy).

Comments: Please write a comment in each code block that explains your thoughts and intentions. Your understanding of the task and the intermediate results will be assessed, not the code.

این پروژه شامل 2 فایل مهم است، لطفا قبل از ارسال پیشنهاد حتما نسبت به بررسی این فایل اقدام فرمایید.

مهارت ها و تخصص های مورد نیاز

---

---