def similarity_threshold(similarity_method, similarity_method_name,
similarity_method_name_short):
if similarity_method == rec.cosine_similarity:
prediction_method = rec.resnicks_cosine_prediction
else:
prediction_method = rec.resnicks_pearson_prediction
print(
"---------------------------------------------------------------------------"
)
print(
"* Please be patient. Generating predictions can take a few minutes.\n"
)
print("Generating predictions...")
print("Neighborhoods created by:\tMinimum similarity threshold")
print("Similarity metric used: \t", similarity_method_name, "\n")
min_corated_items = 5
user_similarity, user_sim_mean, user_sim_stdev = rec.populate_user_similarity(
similarity_method, min_corated_items, users)
similarity_min_cutoffs = [0.95, 0.90, 0.80, 0.70, 0.60, 0.50]
if similarity_method_name == "Pearson's similarity":
similarity_min_cutoffs = [
cos_range_to_pearson_range(s) for s in similarity_min_cutoffs
]
rmse_all = []
time_all = []
coverage_all = []
print(
"\nGenerating predictions across different minimum similarity thresholds for purposes of comparison:"
)
for min_similarity_req in similarity_min_cutoffs:
print("- Minimum similarity threshold: ", min_similarity_req)
# Create neighborhood
rec.assign_neighbors_threshold(users, user_similarity,
min_similarity_req)
# Make predictions
output_file_name = "resnicks_" + similarity_method_name_short + str(
min_similarity_req) + "_min_results.csv"
overall_rmse, total_time, coverage = rec.make_predictions(
prediction_method, output_file_name, users, user_similarity,
item_count, user_count)
print("\tResults saved in:\t", output_file_name)
rmse_all.append(overall_rmse)
time_all.append(total_time)
coverage_all.append(coverage)
curve_label = "Resnick's Formula with " + similarity_method_name
generate_graphs(
similarity_min_cutoffs, rmse_all, time_all, coverage_all, curve_label,
"resnicks_" + similarity_method_name_short + "min_similarity_",
"Minimum Similarity Between Neighbors")
print(
"---------------------------------------------------------------------------"
)
return
def top_percent(similarity_method, similarity_method_name,
similarity_method_name_short):
if similarity_method == rec.cosine_similarity:
prediction_method = rec.resnicks_cosine_prediction
else:
prediction_method = rec.resnicks_pearson_prediction
print(
"---------------------------------------------------------------------------"
)
print(
"* Please be patient. Generating predictions can take a few minutes.\n"
)
print("Generating predictions...")
print("Neighborhoods created by:\tTop n percent of closest users")
print("Similarity metric used: \t", similarity_method_name, "\n")
min_corated_items = 5
percent = [0.025, .05, .10, .15, .20, .25]
rmse_all = []
time_all = []
coverage_all = []
user_similarity, user_sim_mean, user_sim_stdev = rec.populate_user_similarity(
similarity_method, min_corated_items, users)
print(
"\nGenerating predictions across different top percents for purposes of comparison:"
)
for p in percent:
print("Top", p, "% of closest users as neighbors")
# Assign neighborhoods
rec.assign_neighbors_percent(users, user_similarity, p,
similarity_method)
output_file_name = "resnicks_" + similarity_method_name_short + str(
p) + "_percent_results.csv"
overall_rmse, total_time, coverage = rec.make_predictions(
prediction_method, output_file_name, users, user_similarity,
item_count, user_count)
print("\tResults saved in:\t", output_file_name)
rmse_all.append(overall_rmse)
time_all.append(total_time)
coverage_all.append(coverage)
curve_label = "Resnick's Formula with " + similarity_method_name
generate_graphs(
percent, rmse_all, time_all, coverage_all, curve_label,
"resnicks_" + similarity_method_name_short + "top_percent_",
"Top N% of Neighbors")
print(
"---------------------------------------------------------------------------"
)
def k_nearest(similarity_method, similarity_method_name,
similarity_method_name_short):
if similarity_method == rec.cosine_similarity:
prediction_method = rec.resnicks_cosine_prediction
else:
prediction_method = rec.resnicks_pearson_prediction
print(
"---------------------------------------------------------------------------"
)
print(
"* Please be patient. Generating predictions can take a few minutes.\n"
)
print("Generating predictions...")
print("Neighborhoods created by:\tK-nearest users")
print("Similarity metric used: \t", similarity_method_name, "\n")
min_corated_items = 5
k_vals = [5, 10, 20, 30, 40, 50, 60]
rmse_all = []
time_all = []
coverage_all = []
user_similarity, user_sim_mean, user_sim_stdev = rec.populate_user_similarity(
similarity_method, min_corated_items, users)
print(
"\nGenerating predictions across different top percents for purposes of comparison:"
)
for k in k_vals:
print(k, "closest users as neighbors")
# Assign neighborhoods
rec.assign_neighbors_k_nearest(users, user_similarity, k,
similarity_method)
output_file_name = "resnicks_" + similarity_method_name_short + str(
k) + "_closest_results.csv"
# Make predictions
overall_rmse, total_time, coverage = rec.make_predictions(
prediction_method, output_file_name, users, user_similarity,
item_count, user_count)
print("\tResults saved in:\t", output_file_name)
rmse_all.append(overall_rmse)
time_all.append(total_time)
coverage_all.append(coverage)
curve_label = "Resnick's Formula with " + similarity_method_name
generate_graphs(k_vals, rmse_all, time_all, coverage_all, curve_label,
"resnicks_" + similarity_method_name_short + "k_nearest_",
"K Nearest Neighbors")
print(
"---------------------------------------------------------------------------"
)